diff --git a/README.md b/README.md index 2e5feda..ddd89c5 100644 --- a/README.md +++ b/README.md @@ -36,9 +36,9 @@ CI benchmark history from AMD Radeon 8060S (Strix Halo). Detailed data is in `be | Model | Bits | Prompt TPS | Generation TPS | Peak memory (GB) | mlx-vulkan | mlx | Run | | --- | --- | ---: | ---: | ---: | --- | --- | --- | -| mlx-community/Qwen3-0.6B-8bit | 8bit | 4137.078 | 119.812 | 1.997 | 371c566 | 430a327 | [run](https://github.com/goniz/mlx-vulkan/actions/runs/29475951206) | -| mlx-community/Qwen3-0.6B-bf16 | bf16 | 4178.913 | 68.035 | 2.254 | 371c566 | 430a327 | [run](https://github.com/goniz/mlx-vulkan/actions/runs/29475951206) | -| mlx-community/Qwen3.6-35B-A3B-8bit | 8bit | 851.294 | 29.863 | 39.886 | 371c566 | 430a327 | [run](https://github.com/goniz/mlx-vulkan/actions/runs/29475951206) | +| mlx-community/Qwen3-0.6B-8bit | 8bit | 4124.938 | 119.797 | 1.998 | e6b2da3 | ea5289c | [run](https://github.com/goniz/mlx-vulkan/actions/runs/29495909657) | +| mlx-community/Qwen3-0.6B-bf16 | bf16 | 4178.176 | 67.994 | 2.262 | e6b2da3 | ea5289c | [run](https://github.com/goniz/mlx-vulkan/actions/runs/29495909657) | +| mlx-community/Qwen3.6-35B-A3B-8bit | 8bit | 903.405 | 29.568 | 39.885 | e6b2da3 | ea5289c | [run](https://github.com/goniz/mlx-vulkan/actions/runs/29495909657) | ### Model Generation Report @@ -49,10 +49,10 @@ Serial generation smoke tests validate that each model produces coherent output | mlx-community/Qwen3-0.6B-bf16 | pass | pass | 1.143 | Okay, the user wants a concise sentence about why Vulkan acceleration is useful. Let... | | | mlx-community/Qwen3-0.6B-8bit | pass | pass | 0.624 | Okay, the user wants a concise sentence about why Vulkan acceleration is useful. Let... | | | LiquidAI/LFM2.5-1.2B-Instruct-MLX-8bit | pass | pass | 1.183 | Vulkan acceleration enhances performance by enabling efficient parallel processing and reduci... | | -| mlx-community/Qwen3.5-2B-bf16 | pass | pass | 3.544 | Thinking Process: 1. **Analyze the Request:** * Task: Write one concise sentence. * Topic: Wh... | | +| mlx-community/Qwen3.5-2B-bf16 | pass | pass | 3.541 | Thinking Process: 1. **Analyze the Request:** * Task: Write one concise sentence. * Topic: Wh... | | | mlx-community/gemma-4-e2b-it-bf16 | pass | pass | 8.639 | <\|channel>thoughting process:1. **Analyze Request:** The user wants "one concise sentence" ex... | | -| mlx-community/gemma-4-e4b-it-4bit | pass | pass | 3.967 | <\|channel>thought 1. **Analyze the request:** The user wants *one concise sentence* explainin... | | +| mlx-community/gemma-4-e4b-it-4bit | pass | pass | 3.968 | <\|channel>thought 1. **Analyze the request:** The user wants *one concise sentence* explainin... | | | mlx-community/gemma-4-26b-a4b-it-4bit | pass | pass | 13.308 | <\|channel>thought * Topic: Why Vulkan acceleration is useful. * Constraint: One concise sente... | | -| mlx-community/Qwen3.6-35B-A3B-8bit | pass | pass | 34.413 | Here's a thinking process: 1. **Analyze User Input:** - **Topic:** Vulkan acceleration - **Re... | | +| mlx-community/Qwen3.6-35B-A3B-8bit | pass | pass | 34.41 | Here's a thinking process: 1. **Analyze User Input:** - **Topic:** Vulkan acceleration - **Re... | | | mlx-community/gpt-oss-20b-MXFP4-Q8 | pass | pass | 11.436 | <\|channel\|>analysis<\|message\|>We need to write one concise sentence about why Vulkan accelera... | | -| mlx-community/Qwen3.6-27B-8bit | pass | pass | 26.888 | Here's a thinking process: 1. **Analyze User Input:** - **Topic:** Vulkan acceleration - **Re... | | +| mlx-community/Qwen3.6-27B-8bit | pass | pass | 26.871 | Here's a thinking process: 1. **Analyze User Input:** - **Topic:** Vulkan acceleration - **Re... | | diff --git a/benchmarks/README.md b/benchmarks/README.md index 043050a..d50ed0b 100644 --- a/benchmarks/README.md +++ b/benchmarks/README.md @@ -24,9 +24,9 @@ Benchmark data is stored in `results.csv`. Graphs are regenerated from that file | Model | Bits | Prompt TPS | Generation TPS | Peak memory (GB) | mlx-vulkan | mlx | Run | | --- | --- | ---: | ---: | ---: | --- | --- | --- | -| mlx-community/Qwen3-0.6B-8bit | 8bit | 4137.078 | 119.812 | 1.997 | 371c566 | 430a327 | [run](https://github.com/goniz/mlx-vulkan/actions/runs/29475951206) | -| mlx-community/Qwen3-0.6B-bf16 | bf16 | 4178.913 | 68.035 | 2.254 | 371c566 | 430a327 | [run](https://github.com/goniz/mlx-vulkan/actions/runs/29475951206) | -| mlx-community/Qwen3.6-35B-A3B-8bit | 8bit | 851.294 | 29.863 | 39.886 | 371c566 | 430a327 | [run](https://github.com/goniz/mlx-vulkan/actions/runs/29475951206) | +| mlx-community/Qwen3-0.6B-8bit | 8bit | 4124.938 | 119.797 | 1.998 | e6b2da3 | ea5289c | [run](https://github.com/goniz/mlx-vulkan/actions/runs/29495909657) | +| mlx-community/Qwen3-0.6B-bf16 | bf16 | 4178.176 | 67.994 | 2.262 | e6b2da3 | ea5289c | [run](https://github.com/goniz/mlx-vulkan/actions/runs/29495909657) | +| mlx-community/Qwen3.6-35B-A3B-8bit | 8bit | 903.405 | 29.568 | 39.885 | e6b2da3 | ea5289c | [run](https://github.com/goniz/mlx-vulkan/actions/runs/29495909657) | ## Model Generation Report @@ -37,10 +37,10 @@ Generation smoke tests run with `scripts/model_generation_report.py` through `./ | mlx-community/Qwen3-0.6B-bf16 | pass | pass | 1.143 | Okay, the user wants a concise sentence about why Vulkan acceleration is useful. Let... | | | mlx-community/Qwen3-0.6B-8bit | pass | pass | 0.624 | Okay, the user wants a concise sentence about why Vulkan acceleration is useful. Let... | | | LiquidAI/LFM2.5-1.2B-Instruct-MLX-8bit | pass | pass | 1.183 | Vulkan acceleration enhances performance by enabling efficient parallel processing and reduci... | | -| mlx-community/Qwen3.5-2B-bf16 | pass | pass | 3.544 | Thinking Process: 1. **Analyze the Request:** * Task: Write one concise sentence. * Topic: Wh... | | +| mlx-community/Qwen3.5-2B-bf16 | pass | pass | 3.541 | Thinking Process: 1. **Analyze the Request:** * Task: Write one concise sentence. * Topic: Wh... | | | mlx-community/gemma-4-e2b-it-bf16 | pass | pass | 8.639 | <\|channel>thoughting process:1. **Analyze Request:** The user wants "one concise sentence" ex... | | -| mlx-community/gemma-4-e4b-it-4bit | pass | pass | 3.967 | <\|channel>thought 1. **Analyze the request:** The user wants *one concise sentence* explainin... | | +| mlx-community/gemma-4-e4b-it-4bit | pass | pass | 3.968 | <\|channel>thought 1. **Analyze the request:** The user wants *one concise sentence* explainin... | | | mlx-community/gemma-4-26b-a4b-it-4bit | pass | pass | 13.308 | <\|channel>thought * Topic: Why Vulkan acceleration is useful. * Constraint: One concise sente... | | -| mlx-community/Qwen3.6-35B-A3B-8bit | pass | pass | 34.413 | Here's a thinking process: 1. **Analyze User Input:** - **Topic:** Vulkan acceleration - **Re... | | +| mlx-community/Qwen3.6-35B-A3B-8bit | pass | pass | 34.41 | Here's a thinking process: 1. **Analyze User Input:** - **Topic:** Vulkan acceleration - **Re... | | | mlx-community/gpt-oss-20b-MXFP4-Q8 | pass | pass | 11.436 | <\|channel\|>analysis<\|message\|>We need to write one concise sentence about why Vulkan accelera... | | -| mlx-community/Qwen3.6-27B-8bit | pass | pass | 26.888 | Here's a thinking process: 1. **Analyze User Input:** - **Topic:** Vulkan acceleration - **Re... | | +| mlx-community/Qwen3.6-27B-8bit | pass | pass | 26.871 | Here's a thinking process: 1. **Analyze User Input:** - **Topic:** Vulkan acceleration - **Re... | | diff --git a/benchmarks/generation_tps.svg b/benchmarks/generation_tps.svg index 928f114..2a33142 100644 --- a/benchmarks/generation_tps.svg +++ b/benchmarks/generation_tps.svg @@ -21,8 +21,8 @@ 2026-05-28 14:42 2026-06-18 15:23 2026-07-09 11:58 -2026-07-16 06:13 - +2026-07-16 11:53 + Qwen3-0.6B-8bit: 20.398 Qwen3-0.6B-8bit: 20.660 Qwen3-0.6B-8bit: 18.662 @@ -58,7 +58,7 @@ Qwen3-0.6B-8bit: 90.037 Qwen3-0.6B-8bit: 90.042 Qwen3-0.6B-8bit: 121.007 -Qwen3-0.6B-8bit: 119.812 +Qwen3-0.6B-8bit: 119.797 Qwen3-0.6B-8bit @@ -97,7 +97,7 @@ Qwen3-0.6B-bf16: 66.848 Qwen3-0.6B-bf16: 67.037 Qwen3-0.6B-bf16: 67.981 -Qwen3-0.6B-bf16: 68.035 +Qwen3-0.6B-bf16: 67.994 Qwen3-0.6B-bf16 diff --git a/benchmarks/generation_tps_qwen3_6_35b_a3b.svg b/benchmarks/generation_tps_qwen3_6_35b_a3b.svg index a474ed6..88cae2f 100644 --- a/benchmarks/generation_tps_qwen3_6_35b_a3b.svg +++ b/benchmarks/generation_tps_qwen3_6_35b_a3b.svg @@ -4,15 +4,15 @@ 0.0 -6.7 +6.6 -13.4 +13.2 -20.1 +19.9 -26.8 +26.5 -33.4 +33.1 2026-05-28 17:33 @@ -21,25 +21,25 @@ 2026-07-04 21:20 2026-07-09 12:42 2026-07-15 10:47 -2026-07-16 06:13 - -Qwen3.6-35B-A3B-8bit: 8.268 -Qwen3.6-35B-A3B-8bit: 8.022 -Qwen3.6-35B-A3B-8bit: 8.362 -Qwen3.6-35B-A3B-8bit: 21.135 -Qwen3.6-35B-A3B-8bit: 21.113 -Qwen3.6-35B-A3B-8bit: 21.288 -Qwen3.6-35B-A3B-8bit: 20.996 -Qwen3.6-35B-A3B-8bit: 20.969 -Qwen3.6-35B-A3B-8bit: 21.320 -Qwen3.6-35B-A3B-8bit: 21.898 -Qwen3.6-35B-A3B-8bit: 21.956 -Qwen3.6-35B-A3B-8bit: 22.127 -Qwen3.6-35B-A3B-8bit: 22.189 -Qwen3.6-35B-A3B-8bit: 22.187 -Qwen3.6-35B-A3B-8bit: 21.922 -Qwen3.6-35B-A3B-8bit: 29.409 -Qwen3.6-35B-A3B-8bit: 29.863 +2026-07-16 11:53 + +Qwen3.6-35B-A3B-8bit: 8.268 +Qwen3.6-35B-A3B-8bit: 8.022 +Qwen3.6-35B-A3B-8bit: 8.362 +Qwen3.6-35B-A3B-8bit: 21.135 +Qwen3.6-35B-A3B-8bit: 21.113 +Qwen3.6-35B-A3B-8bit: 21.288 +Qwen3.6-35B-A3B-8bit: 20.996 +Qwen3.6-35B-A3B-8bit: 20.969 +Qwen3.6-35B-A3B-8bit: 21.320 +Qwen3.6-35B-A3B-8bit: 21.898 +Qwen3.6-35B-A3B-8bit: 21.956 +Qwen3.6-35B-A3B-8bit: 22.127 +Qwen3.6-35B-A3B-8bit: 22.189 +Qwen3.6-35B-A3B-8bit: 22.187 +Qwen3.6-35B-A3B-8bit: 21.922 +Qwen3.6-35B-A3B-8bit: 29.409 +Qwen3.6-35B-A3B-8bit: 29.568 Qwen3.6-35B-A3B-8bit diff --git a/benchmarks/model_generation_report.json b/benchmarks/model_generation_report.json index 9ec6a30..6084343 100644 --- a/benchmarks/model_generation_report.json +++ b/benchmarks/model_generation_report.json @@ -4,7 +4,7 @@ "generated_output": true, "output_was_coherent": true, "error_msg": "", - "peak_mem_bytes": 1227068592, + "peak_mem_bytes": 1227199716, "peak_mem_gb": 1.143, "output": "\nOkay, the user wants a concise sentence about why Vulkan acceleration is useful. Let me start by recalling what Vulkan is. It's a graphics API developed by NVIDIA, right? It's used for rendering 3D graphics, especially in games and applications that require high performance.\n\nNow, the key points here are" }, @@ -13,7 +13,7 @@ "generated_output": true, "output_was_coherent": true, "error_msg": "", - "peak_mem_bytes": 669668504, + "peak_mem_bytes": 670536904, "peak_mem_gb": 0.624, "output": "\nOkay, the user wants a concise sentence about why Vulkan acceleration is useful. Let me start by recalling what Vulkan is. It's a graphics API developed by NVIDIA, right? It's used for rendering 3D graphics, especially in games and applications that require high performance.\n\nNow, the key points here are" }, @@ -31,8 +31,8 @@ "generated_output": true, "output_was_coherent": true, "error_msg": "", - "peak_mem_bytes": 3805223452, - "peak_mem_gb": 3.544, + "peak_mem_bytes": 3801871168, + "peak_mem_gb": 3.541, "output": "Thinking Process:\n\n1. **Analyze the Request:**\n * Task: Write one concise sentence.\n * Topic: Why Vulkan acceleration is useful.\n * Constraint: Concise (one sentence).\n\n2. **Identify Key Concepts of Vulkan:**\n *" }, { @@ -40,7 +40,7 @@ "generated_output": true, "output_was_coherent": true, "error_msg": "", - "peak_mem_bytes": 9275806172, + "peak_mem_bytes": 9276110256, "peak_mem_gb": 8.639, "output": "<|channel>thoughting process:1. **Analyze Request:** The user wants \"one concise sentence\" explaining about the usefulness of \"Vulkan acceleration.\"\n\n2. **Identify Core Concept (Vulkan Acceleration):** What is Vulkan acceleration? It's a technique (often involving Vul/Vulkan/Wave" }, @@ -49,8 +49,8 @@ "generated_output": true, "output_was_coherent": true, "error_msg": "", - "peak_mem_bytes": 4259721152, - "peak_mem_gb": 3.967, + "peak_mem_bytes": 4260315812, + "peak_mem_gb": 3.968, "output": "<|channel>thought\n1. **Analyze the request:** The user wants *one concise sentence* explaining the usefulness of Vulkan acceleration.\n2. **Identify the core function of Vulkan:** Vulkan is a modern, low-overhead, high-performance graphics and compute API. Its primary benefit is giving developers" }, { @@ -58,7 +58,7 @@ "generated_output": true, "output_was_coherent": true, "error_msg": "", - "peak_mem_bytes": 14289845528, + "peak_mem_bytes": 14289846808, "peak_mem_gb": 13.308, "output": "<|channel>thought\n* Topic: Why Vulkan acceleration is useful.\n * Constraint: One concise sentence.\n\n * Low-level API.\n * Reduced CPU overhead.\n * Better multi-threading/parallelism.\n * Direct hardware access.\n *" }, @@ -67,8 +67,8 @@ "generated_output": true, "output_was_coherent": true, "error_msg": "", - "peak_mem_bytes": 36951096486, - "peak_mem_gb": 34.413, + "peak_mem_bytes": 36947418244, + "peak_mem_gb": 34.41, "output": "Here's a thinking process:\n\n1. **Analyze User Input:**\n - **Topic:** Vulkan acceleration\n - **Requirement:** Write one concise sentence about why it's useful\n - **Key constraints:** One sentence, concise, explains usefulness\n\n2. **Understand Vulkan Acceleration:**" }, { @@ -85,8 +85,8 @@ "generated_output": true, "output_was_coherent": true, "error_msg": "", - "peak_mem_bytes": 28870538098, - "peak_mem_gb": 26.888, + "peak_mem_bytes": 28852616060, + "peak_mem_gb": 26.871, "output": "Here's a thinking process:\n\n1. **Analyze User Input:**\n - **Topic:** Vulkan acceleration\n - **Requirement:** One concise sentence\n - **Focus:** Why it's useful\n\n2. **Identify Key Concepts of Vulkan:**\n - Low-level graphics/compute API" } ] diff --git a/benchmarks/prompt_tps.svg b/benchmarks/prompt_tps.svg index 2996d1e..9738492 100644 --- a/benchmarks/prompt_tps.svg +++ b/benchmarks/prompt_tps.svg @@ -4,15 +4,15 @@ 0.0 -936.1 +935.9 -1872.2 +1871.8 -2808.2 +2807.7 -3744.3 +3743.6 -4680.4 +4679.6 2026-05-18 08:58 @@ -21,8 +21,8 @@ 2026-05-28 14:42 2026-06-18 15:23 2026-07-09 11:58 -2026-07-16 06:13 - +2026-07-16 11:53 + Qwen3-0.6B-8bit: 1317.353 Qwen3-0.6B-8bit: 1324.214 Qwen3-0.6B-8bit: 1302.198 @@ -31,9 +31,9 @@ Qwen3-0.6B-8bit: 1266.470 Qwen3-0.6B-8bit: 1313.222 Qwen3-0.6B-8bit: 1317.869 -Qwen3-0.6B-8bit: 1304.978 +Qwen3-0.6B-8bit: 1304.978 Qwen3-0.6B-8bit: 1320.984 -Qwen3-0.6B-8bit: 1311.879 +Qwen3-0.6B-8bit: 1311.879 Qwen3-0.6B-8bit: 1328.602 Qwen3-0.6B-8bit: 1320.912 Qwen3-0.6B-8bit: 1329.371 @@ -41,63 +41,63 @@ Qwen3-0.6B-8bit: 1302.833 Qwen3-0.6B-8bit: 1331.303 Qwen3-0.6B-8bit: 1329.738 -Qwen3-0.6B-8bit: 1336.072 +Qwen3-0.6B-8bit: 1336.072 Qwen3-0.6B-8bit: 1336.223 Qwen3-0.6B-8bit: 1319.221 Qwen3-0.6B-8bit: 1344.868 Qwen3-0.6B-8bit: 1360.581 Qwen3-0.6B-8bit: 1477.926 Qwen3-0.6B-8bit: 1464.916 -Qwen3-0.6B-8bit: 1458.232 +Qwen3-0.6B-8bit: 1458.232 Qwen3-0.6B-8bit: 1470.973 Qwen3-0.6B-8bit: 1493.533 Qwen3-0.6B-8bit: 1492.104 Qwen3-0.6B-8bit: 1493.524 Qwen3-0.6B-8bit: 1493.076 Qwen3-0.6B-8bit: 1497.589 -Qwen3-0.6B-8bit: 2788.160 -Qwen3-0.6B-8bit: 2820.932 -Qwen3-0.6B-8bit: 2882.832 -Qwen3-0.6B-8bit: 4137.078 +Qwen3-0.6B-8bit: 2788.160 +Qwen3-0.6B-8bit: 2820.932 +Qwen3-0.6B-8bit: 2882.832 +Qwen3-0.6B-8bit: 4124.938 Qwen3-0.6B-8bit - + Qwen3-0.6B-bf16: 2295.900 Qwen3-0.6B-bf16: 2302.213 -Qwen3-0.6B-bf16: 2282.178 +Qwen3-0.6B-bf16: 2282.178 Qwen3-0.6B-bf16: 2263.796 -Qwen3-0.6B-bf16: 2337.462 +Qwen3-0.6B-bf16: 2337.462 Qwen3-0.6B-bf16: 2300.015 -Qwen3-0.6B-bf16: 2270.406 -Qwen3-0.6B-bf16: 2289.089 +Qwen3-0.6B-bf16: 2270.406 +Qwen3-0.6B-bf16: 2289.089 Qwen3-0.6B-bf16: 2257.811 Qwen3-0.6B-bf16: 2286.532 Qwen3-0.6B-bf16: 2264.269 Qwen3-0.6B-bf16: 2315.715 Qwen3-0.6B-bf16: 2293.349 -Qwen3-0.6B-bf16: 2335.527 +Qwen3-0.6B-bf16: 2335.527 Qwen3-0.6B-bf16: 2348.431 Qwen3-0.6B-bf16: 2334.922 Qwen3-0.6B-bf16: 2343.234 -Qwen3-0.6B-bf16: 2332.092 +Qwen3-0.6B-bf16: 2332.092 Qwen3-0.6B-bf16: 2349.782 -Qwen3-0.6B-bf16: 2346.036 +Qwen3-0.6B-bf16: 2346.036 Qwen3-0.6B-bf16: 2392.984 Qwen3-0.6B-bf16: 2333.071 Qwen3-0.6B-bf16: 2410.127 -Qwen3-0.6B-bf16: 2860.579 +Qwen3-0.6B-bf16: 2860.579 Qwen3-0.6B-bf16: 2860.080 -Qwen3-0.6B-bf16: 2862.407 +Qwen3-0.6B-bf16: 2862.407 Qwen3-0.6B-bf16: 2818.647 -Qwen3-0.6B-bf16: 2841.434 -Qwen3-0.6B-bf16: 2812.254 +Qwen3-0.6B-bf16: 2841.434 +Qwen3-0.6B-bf16: 2812.254 Qwen3-0.6B-bf16: 2833.533 Qwen3-0.6B-bf16: 2825.530 Qwen3-0.6B-bf16: 2819.923 -Qwen3-0.6B-bf16: 2839.983 +Qwen3-0.6B-bf16: 2839.983 Qwen3-0.6B-bf16: 2812.893 Qwen3-0.6B-bf16: 2949.229 -Qwen3-0.6B-bf16: 4178.913 +Qwen3-0.6B-bf16: 4178.176 Qwen3-0.6B-bf16 diff --git a/benchmarks/prompt_tps_qwen3_6_35b_a3b.svg b/benchmarks/prompt_tps_qwen3_6_35b_a3b.svg index 15c3a11..c07d398 100644 --- a/benchmarks/prompt_tps_qwen3_6_35b_a3b.svg +++ b/benchmarks/prompt_tps_qwen3_6_35b_a3b.svg @@ -4,15 +4,15 @@ 0.0 -190.7 +202.4 -381.4 +404.7 -572.1 +607.1 -762.8 +809.5 -953.4 +1011.8 2026-05-28 17:33 @@ -21,25 +21,25 @@ 2026-07-04 21:20 2026-07-09 12:42 2026-07-15 10:47 -2026-07-16 06:13 - -Qwen3.6-35B-A3B-8bit: 56.402 -Qwen3.6-35B-A3B-8bit: 54.350 -Qwen3.6-35B-A3B-8bit: 57.111 -Qwen3.6-35B-A3B-8bit: 121.011 -Qwen3.6-35B-A3B-8bit: 124.666 -Qwen3.6-35B-A3B-8bit: 120.009 -Qwen3.6-35B-A3B-8bit: 120.704 -Qwen3.6-35B-A3B-8bit: 127.088 -Qwen3.6-35B-A3B-8bit: 129.037 -Qwen3.6-35B-A3B-8bit: 126.534 -Qwen3.6-35B-A3B-8bit: 125.717 -Qwen3.6-35B-A3B-8bit: 127.692 -Qwen3.6-35B-A3B-8bit: 127.663 -Qwen3.6-35B-A3B-8bit: 148.147 -Qwen3.6-35B-A3B-8bit: 526.499 -Qwen3.6-35B-A3B-8bit: 683.012 -Qwen3.6-35B-A3B-8bit: 851.294 +2026-07-16 11:53 + +Qwen3.6-35B-A3B-8bit: 56.402 +Qwen3.6-35B-A3B-8bit: 54.350 +Qwen3.6-35B-A3B-8bit: 57.111 +Qwen3.6-35B-A3B-8bit: 121.011 +Qwen3.6-35B-A3B-8bit: 124.666 +Qwen3.6-35B-A3B-8bit: 120.009 +Qwen3.6-35B-A3B-8bit: 120.704 +Qwen3.6-35B-A3B-8bit: 127.088 +Qwen3.6-35B-A3B-8bit: 129.037 +Qwen3.6-35B-A3B-8bit: 126.534 +Qwen3.6-35B-A3B-8bit: 125.717 +Qwen3.6-35B-A3B-8bit: 127.692 +Qwen3.6-35B-A3B-8bit: 127.663 +Qwen3.6-35B-A3B-8bit: 148.147 +Qwen3.6-35B-A3B-8bit: 526.499 +Qwen3.6-35B-A3B-8bit: 683.012 +Qwen3.6-35B-A3B-8bit: 903.405 Qwen3.6-35B-A3B-8bit diff --git a/benchmarks/results.csv b/benchmarks/results.csv index f4be676..408ab6b 100644 --- a/benchmarks/results.csv +++ b/benchmarks/results.csv @@ -1,90 +1,93 @@ -timestamp_utc,github_run_id,github_run_number,github_run_attempt,github_workflow,ci_event,ci_ref,run_url,mlx_vulkan_commit,mlx_commit,model_name,model_bits,prompt_tokens,generation_tokens,batch_size,prompt_tps,generation_tps,peak_memory_gb -2026-05-18T08:58:24Z,,,,,legacy-readme,main,,5bdfd05f7804ce0a84e37c129330f9c377d13289,49f9a1df3917963752ba0cd299c8b9f430f4538a,mlx-community/Qwen3-0.6B-bf16,bf16,4096,128,1,2295.900,33.417,2.825 -2026-05-18T08:58:24Z,,,,,legacy-readme,main,,5bdfd05f7804ce0a84e37c129330f9c377d13289,49f9a1df3917963752ba0cd299c8b9f430f4538a,mlx-community/Qwen3-0.6B-8bit,8bit,4096,128,1,1317.353,20.398,2.823 -2026-05-18T11:10:19Z,,,,,local,,,5bdfd05f7804ce0a84e37c129330f9c377d13289,49f9a1df3917963752ba0cd299c8b9f430f4538a,mlx-community/Qwen3-0.6B-bf16,bf16,4096,128,1,2302.213,34.386,2.825 -2026-05-18T11:10:19Z,,,,,local,,,5bdfd05f7804ce0a84e37c129330f9c377d13289,49f9a1df3917963752ba0cd299c8b9f430f4538a,mlx-community/Qwen3-0.6B-8bit,8bit,4096,128,1,1324.214,20.660,2.823 -2026-05-18T11:40:12Z,26031178686,22,1,Vulkan Build and Benchmark,push,main,https://github.com/goniz/mlx-vulkan/actions/runs/26031178686,db88189558430aca509d9e15bb8a8f0a08ac83ef,49f9a1df3917963752ba0cd299c8b9f430f4538a,mlx-community/Qwen3-0.6B-bf16,bf16,4096,128,1,2282.178,33.745,2.825 -2026-05-18T11:40:12Z,26031178686,22,1,Vulkan Build and Benchmark,push,main,https://github.com/goniz/mlx-vulkan/actions/runs/26031178686,db88189558430aca509d9e15bb8a8f0a08ac83ef,49f9a1df3917963752ba0cd299c8b9f430f4538a,mlx-community/Qwen3-0.6B-8bit,8bit,4096,128,1,1302.198,18.662,2.823 -2026-05-19T17:57:06Z,26115334542,23,1,Vulkan Build and Benchmark,push,main,https://github.com/goniz/mlx-vulkan/actions/runs/26115334542,13821383e55fd9e88b693dd690565878e43a6bc9,f81649a8cfef762d21eb879149e663ce3dd8b794,mlx-community/Qwen3-0.6B-bf16,bf16,4096,128,1,2263.796,28.465,2.565 -2026-05-19T17:57:06Z,26115334542,23,1,Vulkan Build and Benchmark,push,main,https://github.com/goniz/mlx-vulkan/actions/runs/26115334542,13821383e55fd9e88b693dd690565878e43a6bc9,f81649a8cfef762d21eb879149e663ce3dd8b794,mlx-community/Qwen3-0.6B-8bit,8bit,4096,128,1,1316.035,19.709,1.917 -2026-05-20T05:17:15Z,26142834214,24,1,Vulkan Build and Benchmark,push,main,https://github.com/goniz/mlx-vulkan/actions/runs/26142834214,53a6764931412934649ade16840142594b933982,2a18b7c7a978764ec3cf606b28edc9f991dd4677,mlx-community/Qwen3-0.6B-bf16,bf16,4096,128,1,2337.462,28.049,2.565 -2026-05-20T05:17:15Z,26142834214,24,1,Vulkan Build and Benchmark,push,main,https://github.com/goniz/mlx-vulkan/actions/runs/26142834214,53a6764931412934649ade16840142594b933982,2a18b7c7a978764ec3cf606b28edc9f991dd4677,mlx-community/Qwen3-0.6B-8bit,8bit,4096,128,1,1330.239,19.969,1.917 -2026-05-20T09:39:03Z,26154299689,25,1,Vulkan Build and Benchmark,push,main,https://github.com/goniz/mlx-vulkan/actions/runs/26154299689,33eb2d2d013ebd151cf36c7ac841c70649cb5905,13c309699abde91be39f88b2b475fac9e3251412,mlx-community/Qwen3-0.6B-bf16,bf16,4096,128,1,2300.015,27.642,2.565 -2026-05-20T09:39:03Z,26154299689,25,1,Vulkan Build and Benchmark,push,main,https://github.com/goniz/mlx-vulkan/actions/runs/26154299689,33eb2d2d013ebd151cf36c7ac841c70649cb5905,13c309699abde91be39f88b2b475fac9e3251412,mlx-community/Qwen3-0.6B-8bit,8bit,4096,128,1,1266.470,17.653,1.917 -2026-05-20T10:17:07Z,26156115134,26,1,Vulkan Build and Benchmark,push,main,https://github.com/goniz/mlx-vulkan/actions/runs/26156115134,24c9e5a63ceaea177c13a9549ca117256663a6a6,37d3e21c98c25d9856ee12a2c0116b5df73bcfba,mlx-community/Qwen3-0.6B-bf16,bf16,4096,128,1,2270.406,28.463,2.565 -2026-05-20T10:17:07Z,26156115134,26,1,Vulkan Build and Benchmark,push,main,https://github.com/goniz/mlx-vulkan/actions/runs/26156115134,24c9e5a63ceaea177c13a9549ca117256663a6a6,37d3e21c98c25d9856ee12a2c0116b5df73bcfba,mlx-community/Qwen3-0.6B-8bit,8bit,4096,128,1,1313.222,18.601,1.917 -2026-05-20T17:03:01Z,26177458763,27,1,Vulkan Build and Benchmark,push,main,https://github.com/goniz/mlx-vulkan/actions/runs/26177458763,59b4efefffa5550badca78ceb86bfc7750e7f17b,7e13fc8fdb23c833129aed7f8a2ced8abf119f15,mlx-community/Qwen3-0.6B-bf16,bf16,4096,128,1,2289.089,28.488,2.565 -2026-05-20T17:03:01Z,26177458763,27,1,Vulkan Build and Benchmark,push,main,https://github.com/goniz/mlx-vulkan/actions/runs/26177458763,59b4efefffa5550badca78ceb86bfc7750e7f17b,7e13fc8fdb23c833129aed7f8a2ced8abf119f15,mlx-community/Qwen3-0.6B-8bit,8bit,4096,128,1,1317.869,18.364,1.919 -2026-05-20T17:44:14Z,26179663920,28,1,Vulkan Build and Benchmark,push,main,https://github.com/goniz/mlx-vulkan/actions/runs/26179663920,96fe5ce9a63b5b020bb66660a5718bf9a2cf6d84,e26ac9ffde2f6d9ca4638dd3c9ee2e45c67ede1b,mlx-community/Qwen3-0.6B-bf16,bf16,4096,128,1,2257.811,27.349,2.565 -2026-05-20T17:44:14Z,26179663920,28,1,Vulkan Build and Benchmark,push,main,https://github.com/goniz/mlx-vulkan/actions/runs/26179663920,96fe5ce9a63b5b020bb66660a5718bf9a2cf6d84,e26ac9ffde2f6d9ca4638dd3c9ee2e45c67ede1b,mlx-community/Qwen3-0.6B-8bit,8bit,4096,128,1,1304.978,18.946,1.917 -2026-05-24T18:08:49Z,26368847068,29,1,Vulkan Build and Benchmark,push,main,https://github.com/goniz/mlx-vulkan/actions/runs/26368847068,545e44c9fcb1e63f8776f8d05ebd44fed0a26fb3,045d17b6515c358c3e2cd9630216cb26995474ae,mlx-community/Qwen3-0.6B-bf16,bf16,4096,128,1,2286.532,28.557,2.603 -2026-05-24T18:08:49Z,26368847068,29,1,Vulkan Build and Benchmark,push,main,https://github.com/goniz/mlx-vulkan/actions/runs/26368847068,545e44c9fcb1e63f8776f8d05ebd44fed0a26fb3,045d17b6515c358c3e2cd9630216cb26995474ae,mlx-community/Qwen3-0.6B-8bit,8bit,4096,128,1,1320.984,18.960,1.896 -2026-05-24T19:50:46Z,26371082381,30,1,Vulkan Build and Benchmark,push,main,https://github.com/goniz/mlx-vulkan/actions/runs/26371082381,21e14fb3b4e190bad22184ccc3bd21f345e28344,68a4c6b9e3a4830286426fa645c081b327a23532,mlx-community/Qwen3-0.6B-bf16,bf16,4096,128,1,2264.269,28.568,2.603 -2026-05-24T19:50:46Z,26371082381,30,1,Vulkan Build and Benchmark,push,main,https://github.com/goniz/mlx-vulkan/actions/runs/26371082381,21e14fb3b4e190bad22184ccc3bd21f345e28344,68a4c6b9e3a4830286426fa645c081b327a23532,mlx-community/Qwen3-0.6B-8bit,8bit,4096,128,1,1311.879,18.768,1.901 -2026-05-24T20:02:13Z,,,,,local,,,21e14fb3b4e190bad22184ccc3bd21f345e28344,68a4c6b9e3a4830286426fa645c081b327a23532,mlx-community/Qwen3-0.6B-bf16,bf16,4096,128,1,2315.715,28.600,2.603 -2026-05-24T20:02:13Z,,,,,local,,,21e14fb3b4e190bad22184ccc3bd21f345e28344,68a4c6b9e3a4830286426fa645c081b327a23532,mlx-community/Qwen3-0.6B-8bit,8bit,4096,128,1,1328.602,19.271,1.901 -2026-05-24T20:21:37Z,26371767095,31,1,Vulkan Build and Benchmark,push,main,https://github.com/goniz/mlx-vulkan/actions/runs/26371767095,c1403315237c0e8c41d31c03fb4101d7632236d4,68a4c6b9e3a4830286426fa645c081b327a23532,mlx-community/Qwen3-0.6B-bf16,bf16,4096,128,1,2293.349,28.079,2.603 -2026-05-24T20:21:37Z,26371767095,31,1,Vulkan Build and Benchmark,push,main,https://github.com/goniz/mlx-vulkan/actions/runs/26371767095,c1403315237c0e8c41d31c03fb4101d7632236d4,68a4c6b9e3a4830286426fa645c081b327a23532,mlx-community/Qwen3-0.6B-8bit,8bit,4096,128,1,1320.912,18.808,1.898 -2026-05-24T22:57:10Z,26375042081,32,1,Vulkan Build and Benchmark,push,main,https://github.com/goniz/mlx-vulkan/actions/runs/26375042081,ea6f0f476b1ab75ca9c9657303d011b7fc90cff6,8e4a5d8200a0aa8acb02c401b291465b7f2deca8,mlx-community/Qwen3-0.6B-bf16,bf16,4096,128,1,2335.527,62.190,2.824 -2026-05-24T22:57:10Z,26375042081,32,1,Vulkan Build and Benchmark,push,main,https://github.com/goniz/mlx-vulkan/actions/runs/26375042081,ea6f0f476b1ab75ca9c9657303d011b7fc90cff6,8e4a5d8200a0aa8acb02c401b291465b7f2deca8,mlx-community/Qwen3-0.6B-8bit,8bit,4096,128,1,1329.371,25.035,2.204 -2026-05-25T05:38:25Z,26385119043,33,1,Vulkan Build and Benchmark,push,main,https://github.com/goniz/mlx-vulkan/actions/runs/26385119043,514902c9c32f1df2e98b631e79c458585092d0f8,51787f26d9dd78e4e88a117ca45256eb30f441bd,mlx-community/Qwen3-0.6B-bf16,bf16,4096,128,1,2348.431,67.708,2.824 -2026-05-25T05:38:25Z,26385119043,33,1,Vulkan Build and Benchmark,push,main,https://github.com/goniz/mlx-vulkan/actions/runs/26385119043,514902c9c32f1df2e98b631e79c458585092d0f8,51787f26d9dd78e4e88a117ca45256eb30f441bd,mlx-community/Qwen3-0.6B-8bit,8bit,4096,128,1,1326.742,26.004,2.204 -2026-05-25T16:38:02Z,26410543427,34,1,Vulkan Build and Benchmark,push,main,https://github.com/goniz/mlx-vulkan/actions/runs/26410543427,b71d53e623a47597a1412ca379c10247df88774d,aab59b9e2015cc83c891470d1f106ea3b7a831ed,mlx-community/Qwen3-0.6B-bf16,bf16,4096,128,1,2334.922,67.640,2.824 -2026-05-25T16:38:02Z,26410543427,34,1,Vulkan Build and Benchmark,push,main,https://github.com/goniz/mlx-vulkan/actions/runs/26410543427,b71d53e623a47597a1412ca379c10247df88774d,aab59b9e2015cc83c891470d1f106ea3b7a831ed,mlx-community/Qwen3-0.6B-8bit,8bit,4096,128,1,1302.833,26.039,2.204 -2026-05-27T07:30:15Z,26497290169,35,1,Vulkan Build and Benchmark,push,main,https://github.com/goniz/mlx-vulkan/actions/runs/26497290169,68a02370f0e50c00ac7bb8a630a3462286ed88af,ea869877c6948c826beb70ee4d7f5dac3739285f,mlx-community/Qwen3-0.6B-bf16,bf16,4096,128,1,2343.234,63.423,2.824 -2026-05-27T07:30:15Z,26497290169,35,1,Vulkan Build and Benchmark,push,main,https://github.com/goniz/mlx-vulkan/actions/runs/26497290169,68a02370f0e50c00ac7bb8a630a3462286ed88af,ea869877c6948c826beb70ee4d7f5dac3739285f,mlx-community/Qwen3-0.6B-8bit,8bit,4096,128,1,1331.303,28.681,2.204 -2026-05-28T09:45:35Z,26567238160,40,1,Vulkan Build and Benchmark,push,main,https://github.com/goniz/mlx-vulkan/actions/runs/26567238160,d922a89538259f01357550d2b1054415234d542a,38617fc360770cb7f8e448dd828c77f8eb860eab,mlx-community/Qwen3-0.6B-bf16,bf16,4096,128,1,2332.092,62.990,2.824 -2026-05-28T09:45:35Z,26567238160,40,1,Vulkan Build and Benchmark,push,main,https://github.com/goniz/mlx-vulkan/actions/runs/26567238160,d922a89538259f01357550d2b1054415234d542a,38617fc360770cb7f8e448dd828c77f8eb860eab,mlx-community/Qwen3-0.6B-8bit,8bit,4096,128,1,1329.738,27.392,2.242 -2026-05-28T14:42:43Z,26581733030,41,1,Vulkan Build and Benchmark,push,main,https://github.com/goniz/mlx-vulkan/actions/runs/26581733030,00dfd9c235a436d59bb3a327ba70bb455c221ae7,4a6c4a9e66278c18db0e4649423c1e4bee97a279,mlx-community/Qwen3-0.6B-bf16,bf16,4096,128,1,2349.782,63.834,2.824 -2026-05-28T14:42:43Z,26581733030,41,1,Vulkan Build and Benchmark,push,main,https://github.com/goniz/mlx-vulkan/actions/runs/26581733030,00dfd9c235a436d59bb3a327ba70bb455c221ae7,4a6c4a9e66278c18db0e4649423c1e4bee97a279,mlx-community/Qwen3-0.6B-8bit,8bit,4096,128,1,1336.072,27.618,2.204 -2026-05-28T17:33:59Z,26591235365,42,1,Vulkan Build and Benchmark,push,main,https://github.com/goniz/mlx-vulkan/actions/runs/26591235365,de335807c49cd93cd1ded812b48c144c1cbed71a,4a6c4a9e66278c18db0e4649423c1e4bee97a279,mlx-community/Qwen3-0.6B-bf16,bf16,4096,128,1,2346.036,63.854,2.824 -2026-05-28T17:33:59Z,26591235365,42,1,Vulkan Build and Benchmark,push,main,https://github.com/goniz/mlx-vulkan/actions/runs/26591235365,de335807c49cd93cd1ded812b48c144c1cbed71a,4a6c4a9e66278c18db0e4649423c1e4bee97a279,mlx-community/Qwen3-0.6B-8bit,8bit,4096,128,1,1336.223,25.825,2.246 -2026-05-28T17:33:59Z,26591235365,42,1,Vulkan Build and Benchmark,push,main,https://github.com/goniz/mlx-vulkan/actions/runs/26591235365,de335807c49cd93cd1ded812b48c144c1cbed71a,4a6c4a9e66278c18db0e4649423c1e4bee97a279,mlx-community/Qwen3.6-35B-A3B-8bit,8bit,4096,128,1,56.402,8.268,42.442 -2026-05-31T17:12:42Z,26719038679,46,1,Vulkan Build and Benchmark,push,main,https://github.com/goniz/mlx-vulkan/actions/runs/26719038679,8fbf5f9e3d6a2a2d0f16b3cfb086cfb7bbfe815d,0a10dbf9d61295ae2544c763634a590eb2197dd9,mlx-community/Qwen3-0.6B-bf16,bf16,4096,128,1,2392.984,60.641,2.824 -2026-05-31T17:12:42Z,26719038679,46,1,Vulkan Build and Benchmark,push,main,https://github.com/goniz/mlx-vulkan/actions/runs/26719038679,8fbf5f9e3d6a2a2d0f16b3cfb086cfb7bbfe815d,0a10dbf9d61295ae2544c763634a590eb2197dd9,mlx-community/Qwen3-0.6B-8bit,8bit,4096,128,1,1319.221,25.538,2.204 -2026-05-31T17:12:42Z,26719038679,46,1,Vulkan Build and Benchmark,push,main,https://github.com/goniz/mlx-vulkan/actions/runs/26719038679,8fbf5f9e3d6a2a2d0f16b3cfb086cfb7bbfe815d,0a10dbf9d61295ae2544c763634a590eb2197dd9,mlx-community/Qwen3.6-35B-A3B-8bit,8bit,4096,128,1,54.350,8.022,42.438 -2026-06-04T06:13:41Z,26934302484,49,1,Vulkan Build and Benchmark,push,main,https://github.com/goniz/mlx-vulkan/actions/runs/26934302484,23a25d1b8968fa3de6053b11b25cf4a69b97d45c,bc2fdc59c9a3326133aaf1ff6c607c9203ecf47f,mlx-community/Qwen3-0.6B-bf16,bf16,4096,128,1,2333.071,63.860,2.824 -2026-06-04T06:13:41Z,26934302484,49,1,Vulkan Build and Benchmark,push,main,https://github.com/goniz/mlx-vulkan/actions/runs/26934302484,23a25d1b8968fa3de6053b11b25cf4a69b97d45c,bc2fdc59c9a3326133aaf1ff6c607c9203ecf47f,mlx-community/Qwen3-0.6B-8bit,8bit,4096,128,1,1344.868,30.731,2.204 -2026-06-04T06:13:41Z,26934302484,49,1,Vulkan Build and Benchmark,push,main,https://github.com/goniz/mlx-vulkan/actions/runs/26934302484,23a25d1b8968fa3de6053b11b25cf4a69b97d45c,bc2fdc59c9a3326133aaf1ff6c607c9203ecf47f,mlx-community/Qwen3.6-35B-A3B-8bit,8bit,4096,128,1,57.111,8.362,42.440 -2026-06-04T07:22:00Z,26937061971,51,1,Vulkan Build and Benchmark,push,main,https://github.com/goniz/mlx-vulkan/actions/runs/26937061971,c0c3da7aa159b41cb29e469ba481a32843c2350f,5d618c8cecd05b0518721b5450d8e79ad9ab236d,mlx-community/Qwen3-0.6B-bf16,bf16,4096,128,1,2410.127,65.748,2.614 -2026-06-04T07:22:00Z,26937061971,51,1,Vulkan Build and Benchmark,push,main,https://github.com/goniz/mlx-vulkan/actions/runs/26937061971,c0c3da7aa159b41cb29e469ba481a32843c2350f,5d618c8cecd05b0518721b5450d8e79ad9ab236d,mlx-community/Qwen3-0.6B-8bit,8bit,4096,128,1,1360.581,87.609,2.056 -2026-06-04T07:22:00Z,26937061971,51,1,Vulkan Build and Benchmark,push,main,https://github.com/goniz/mlx-vulkan/actions/runs/26937061971,c0c3da7aa159b41cb29e469ba481a32843c2350f,5d618c8cecd05b0518721b5450d8e79ad9ab236d,mlx-community/Qwen3.6-35B-A3B-8bit,8bit,4096,128,1,121.011,21.135,40.350 -2026-06-13T22:02:47Z,27480356548,52,1,Vulkan Build and Benchmark,push,main,https://github.com/goniz/mlx-vulkan/actions/runs/27480356548,d1c89c732e398e2a469bbc206a5292f02f4c677a,75812951eedf0a7f82b1898e56c2281500d2e6c3,mlx-community/Qwen3-0.6B-bf16,bf16,4096,128,1,2860.579,65.949,2.614 -2026-06-13T22:02:47Z,27480356548,52,1,Vulkan Build and Benchmark,push,main,https://github.com/goniz/mlx-vulkan/actions/runs/27480356548,d1c89c732e398e2a469bbc206a5292f02f4c677a,75812951eedf0a7f82b1898e56c2281500d2e6c3,mlx-community/Qwen3-0.6B-8bit,8bit,4096,128,1,1477.926,87.524,2.056 -2026-06-13T22:02:47Z,27480356548,52,1,Vulkan Build and Benchmark,push,main,https://github.com/goniz/mlx-vulkan/actions/runs/27480356548,d1c89c732e398e2a469bbc206a5292f02f4c677a,75812951eedf0a7f82b1898e56c2281500d2e6c3,mlx-community/Qwen3.6-35B-A3B-8bit,8bit,4096,128,1,124.666,21.113,40.349 -2026-06-18T15:23:39Z,27769330530,57,1,Vulkan Build and Benchmark,push,main,https://github.com/goniz/mlx-vulkan/actions/runs/27769330530,4f515907bd714513c82e0426c219804b722a58eb,f799cc26d2f4703872d27b945d6c65ab528b4dee,mlx-community/Qwen3-0.6B-bf16,bf16,4096,128,1,2860.080,65.950,2.614 -2026-06-18T15:23:39Z,27769330530,57,1,Vulkan Build and Benchmark,push,main,https://github.com/goniz/mlx-vulkan/actions/runs/27769330530,4f515907bd714513c82e0426c219804b722a58eb,f799cc26d2f4703872d27b945d6c65ab528b4dee,mlx-community/Qwen3-0.6B-8bit,8bit,4096,128,1,1464.916,87.609,2.056 -2026-06-18T15:23:39Z,27769330530,57,1,Vulkan Build and Benchmark,push,main,https://github.com/goniz/mlx-vulkan/actions/runs/27769330530,4f515907bd714513c82e0426c219804b722a58eb,f799cc26d2f4703872d27b945d6c65ab528b4dee,mlx-community/Qwen3.6-35B-A3B-8bit,8bit,4096,128,1,120.009,21.288,40.350 -2026-06-19T15:37:13Z,,,,,local,,,baaf71e88ab18a1fe4abde2059b4eba48f4ddbb6,f799cc26d2f4703872d27b945d6c65ab528b4dee,mlx-community/Qwen3-0.6B-bf16,bf16,4096,128,1,2862.407,65.486,2.614 -2026-06-19T15:37:13Z,,,,,local,,,baaf71e88ab18a1fe4abde2059b4eba48f4ddbb6,f799cc26d2f4703872d27b945d6c65ab528b4dee,mlx-community/Qwen3-0.6B-8bit,8bit,4096,128,1,1458.232,85.857,2.055 -2026-06-19T15:37:13Z,,,,,local,,,baaf71e88ab18a1fe4abde2059b4eba48f4ddbb6,f799cc26d2f4703872d27b945d6c65ab528b4dee,mlx-community/Qwen3.6-35B-A3B-8bit,8bit,4096,128,1,120.704,20.996,40.349 -2026-07-02T08:24:34Z,28575944585,58,1,Vulkan Build and Benchmark,push,main,https://github.com/goniz/mlx-vulkan/actions/runs/28575944585,28e7e391e6e8a3d7f2afe15827d3fa7203d8a08c,088d06620ed01275ccbff7a16fda30348b563525,mlx-community/Qwen3-0.6B-bf16,bf16,4096,128,1,2818.647,65.027,2.614 -2026-07-02T08:24:34Z,28575944585,58,1,Vulkan Build and Benchmark,push,main,https://github.com/goniz/mlx-vulkan/actions/runs/28575944585,28e7e391e6e8a3d7f2afe15827d3fa7203d8a08c,088d06620ed01275ccbff7a16fda30348b563525,mlx-community/Qwen3-0.6B-8bit,8bit,4096,128,1,1470.973,85.274,2.056 -2026-07-02T08:24:34Z,28575944585,58,1,Vulkan Build and Benchmark,push,main,https://github.com/goniz/mlx-vulkan/actions/runs/28575944585,28e7e391e6e8a3d7f2afe15827d3fa7203d8a08c,088d06620ed01275ccbff7a16fda30348b563525,mlx-community/Qwen3.6-35B-A3B-8bit,8bit,4096,128,1,127.088,20.969,40.349 -2026-07-03T13:26:49Z,28663462997,63,1,Vulkan Build and Benchmark,push,main,https://github.com/goniz/mlx-vulkan/actions/runs/28663462997,ebd8aec5b30fb07ac40dca961277ef96821fe5e0,a00c15f18df4109f4b97a401dfeafe0ea7ce41f2,mlx-community/Qwen3-0.6B-bf16,bf16,4096,128,1,2841.434,62.497,2.723 -2026-07-03T13:26:49Z,28663462997,63,1,Vulkan Build and Benchmark,push,main,https://github.com/goniz/mlx-vulkan/actions/runs/28663462997,ebd8aec5b30fb07ac40dca961277ef96821fe5e0,a00c15f18df4109f4b97a401dfeafe0ea7ce41f2,mlx-community/Qwen3-0.6B-8bit,8bit,4096,128,1,1493.533,83.962,2.165 -2026-07-03T13:26:49Z,28663462997,63,1,Vulkan Build and Benchmark,push,main,https://github.com/goniz/mlx-vulkan/actions/runs/28663462997,ebd8aec5b30fb07ac40dca961277ef96821fe5e0,a00c15f18df4109f4b97a401dfeafe0ea7ce41f2,mlx-community/Qwen3.6-35B-A3B-8bit,8bit,4096,128,1,129.037,21.320,40.039 -2026-07-04T21:20:51Z,28719951837,65,1,Vulkan Build and Benchmark,push,main,https://github.com/goniz/mlx-vulkan/actions/runs/28719951837,f49be611be9644890d3491f70f693828187a7045,e639eb1189d304d6e0e4bc6b6caabc77640fe50e,mlx-community/Qwen3-0.6B-bf16,bf16,4096,128,1,2812.254,66.926,2.101 -2026-07-04T21:20:51Z,28719951837,65,1,Vulkan Build and Benchmark,push,main,https://github.com/goniz/mlx-vulkan/actions/runs/28719951837,f49be611be9644890d3491f70f693828187a7045,e639eb1189d304d6e0e4bc6b6caabc77640fe50e,mlx-community/Qwen3-0.6B-8bit,8bit,4096,128,1,1492.104,89.988,1.554 -2026-07-04T21:20:51Z,28719951837,65,1,Vulkan Build and Benchmark,push,main,https://github.com/goniz/mlx-vulkan/actions/runs/28719951837,f49be611be9644890d3491f70f693828187a7045,e639eb1189d304d6e0e4bc6b6caabc77640fe50e,mlx-community/Qwen3.6-35B-A3B-8bit,8bit,4096,128,1,126.534,21.898,39.235 -2026-07-07T07:47:39Z,28850130432,67,1,Vulkan Build and Benchmark,push,main,https://github.com/goniz/mlx-vulkan/actions/runs/28850130432,61e55afc60f7ba441fa3d1ba1fe4eba4d0384ccf,d2f75a158dd8d16c7f5dd7aced3d598d38767b46,mlx-community/Qwen3-0.6B-bf16,bf16,4096,128,1,2833.533,66.528,2.102 -2026-07-07T07:47:39Z,28850130432,67,1,Vulkan Build and Benchmark,push,main,https://github.com/goniz/mlx-vulkan/actions/runs/28850130432,61e55afc60f7ba441fa3d1ba1fe4eba4d0384ccf,d2f75a158dd8d16c7f5dd7aced3d598d38767b46,mlx-community/Qwen3-0.6B-8bit,8bit,4096,128,1,1493.524,89.920,1.544 -2026-07-07T07:47:39Z,28850130432,67,1,Vulkan Build and Benchmark,push,main,https://github.com/goniz/mlx-vulkan/actions/runs/28850130432,61e55afc60f7ba441fa3d1ba1fe4eba4d0384ccf,d2f75a158dd8d16c7f5dd7aced3d598d38767b46,mlx-community/Qwen3.6-35B-A3B-8bit,8bit,4096,128,1,125.717,21.956,39.235 -2026-07-09T11:58:18Z,29016445983,69,1,Vulkan Build and Benchmark,push,main,https://github.com/goniz/mlx-vulkan/actions/runs/29016445983,8cb61017e9bde777026fde6c4831210fca32ec82,c7c199f79df4b987c134860a92430d660cc312ad,mlx-community/Qwen3-0.6B-bf16,bf16,4096,128,1,2825.530,66.992,2.099 -2026-07-09T11:58:18Z,29016445983,69,1,Vulkan Build and Benchmark,push,main,https://github.com/goniz/mlx-vulkan/actions/runs/29016445983,8cb61017e9bde777026fde6c4831210fca32ec82,c7c199f79df4b987c134860a92430d660cc312ad,mlx-community/Qwen3-0.6B-8bit,8bit,4096,128,1,1493.076,90.933,1.547 -2026-07-09T11:58:18Z,29016445983,69,1,Vulkan Build and Benchmark,push,main,https://github.com/goniz/mlx-vulkan/actions/runs/29016445983,8cb61017e9bde777026fde6c4831210fca32ec82,c7c199f79df4b987c134860a92430d660cc312ad,mlx-community/Qwen3.6-35B-A3B-8bit,8bit,4096,128,1,127.692,22.127,39.235 -2026-07-09T12:42:46Z,29018390272,71,1,Vulkan Build and Benchmark,push,main,https://github.com/goniz/mlx-vulkan/actions/runs/29018390272,6c1d1aa78607ad8c47df6fdc3eb4357f4b319380,4c86a8302d94e33417d45faaef9d551e05d8981b,mlx-community/Qwen3-0.6B-bf16,bf16,4096,128,1,2819.923,66.982,2.103 -2026-07-09T12:42:46Z,29018390272,71,1,Vulkan Build and Benchmark,push,main,https://github.com/goniz/mlx-vulkan/actions/runs/29018390272,6c1d1aa78607ad8c47df6fdc3eb4357f4b319380,4c86a8302d94e33417d45faaef9d551e05d8981b,mlx-community/Qwen3-0.6B-8bit,8bit,4096,128,1,1497.589,89.741,1.552 -2026-07-09T12:42:46Z,29018390272,71,1,Vulkan Build and Benchmark,push,main,https://github.com/goniz/mlx-vulkan/actions/runs/29018390272,6c1d1aa78607ad8c47df6fdc3eb4357f4b319380,4c86a8302d94e33417d45faaef9d551e05d8981b,mlx-community/Qwen3.6-35B-A3B-8bit,8bit,4096,128,1,127.663,22.189,39.283 -2026-07-13T20:09:53Z,29281094972,72,1,Vulkan Build and Benchmark,push,main,https://github.com/goniz/mlx-vulkan/actions/runs/29281094972,fd80e5ae3b752e999fd92588f1acc7f5426ae0ee,277387e9966afe5774da27da5caa9e818a05fb94,mlx-community/Qwen3-0.6B-bf16,bf16,4096,128,1,2839.983,66.848,2.102 -2026-07-13T20:09:53Z,29281094972,72,1,Vulkan Build and Benchmark,push,main,https://github.com/goniz/mlx-vulkan/actions/runs/29281094972,fd80e5ae3b752e999fd92588f1acc7f5426ae0ee,277387e9966afe5774da27da5caa9e818a05fb94,mlx-community/Qwen3-0.6B-8bit,8bit,4096,128,1,2788.160,90.037,1.653 -2026-07-13T20:09:53Z,29281094972,72,1,Vulkan Build and Benchmark,push,main,https://github.com/goniz/mlx-vulkan/actions/runs/29281094972,fd80e5ae3b752e999fd92588f1acc7f5426ae0ee,277387e9966afe5774da27da5caa9e818a05fb94,mlx-community/Qwen3.6-35B-A3B-8bit,8bit,4096,128,1,148.147,22.187,39.245 -2026-07-14T14:36:20Z,29341360981,73,1,Vulkan Build and Benchmark,push,main,https://github.com/goniz/mlx-vulkan/actions/runs/29341360981,58d5f072e9d0dc42aafeac665f299ab685ebbc9a,501c4b5571a1e1be85f11e244473a67123b50465,mlx-community/Qwen3-0.6B-bf16,bf16,4096,128,1,2812.893,67.037,2.099 -2026-07-14T14:36:20Z,29341360981,73,1,Vulkan Build and Benchmark,push,main,https://github.com/goniz/mlx-vulkan/actions/runs/29341360981,58d5f072e9d0dc42aafeac665f299ab685ebbc9a,501c4b5571a1e1be85f11e244473a67123b50465,mlx-community/Qwen3-0.6B-8bit,8bit,4096,128,1,2820.932,90.042,1.671 -2026-07-14T14:36:20Z,29341360981,73,1,Vulkan Build and Benchmark,push,main,https://github.com/goniz/mlx-vulkan/actions/runs/29341360981,58d5f072e9d0dc42aafeac665f299ab685ebbc9a,501c4b5571a1e1be85f11e244473a67123b50465,mlx-community/Qwen3.6-35B-A3B-8bit,8bit,4096,128,1,526.499,21.922,39.260 -2026-07-15T10:47:58Z,29409233711,78,1,Vulkan Build and Benchmark,push,main,https://github.com/goniz/mlx-vulkan/actions/runs/29409233711,76e10ae33b93eec085bc509dd7d8b30be4420af0,df10f7b8fa535bf4b07e6ff6e3082f38565bd20f,mlx-community/Qwen3-0.6B-bf16,bf16,4096,128,1,2949.229,67.981,2.102 -2026-07-15T10:47:58Z,29409233711,78,1,Vulkan Build and Benchmark,push,main,https://github.com/goniz/mlx-vulkan/actions/runs/29409233711,76e10ae33b93eec085bc509dd7d8b30be4420af0,df10f7b8fa535bf4b07e6ff6e3082f38565bd20f,mlx-community/Qwen3-0.6B-8bit,8bit,4096,128,1,2882.832,121.007,1.688 -2026-07-15T10:47:58Z,29409233711,78,1,Vulkan Build and Benchmark,push,main,https://github.com/goniz/mlx-vulkan/actions/runs/29409233711,76e10ae33b93eec085bc509dd7d8b30be4420af0,df10f7b8fa535bf4b07e6ff6e3082f38565bd20f,mlx-community/Qwen3.6-35B-A3B-8bit,8bit,4096,128,1,683.012,29.409,39.665 -2026-07-16T06:13:04Z,29475951206,88,1,Vulkan Build and Benchmark,push,main,https://github.com/goniz/mlx-vulkan/actions/runs/29475951206,371c56655b502f49c021fe8ebea09d4f3ce7d686,430a3270612c8b0de2eec29a0f52c18c4fdc81b9,mlx-community/Qwen3-0.6B-bf16,bf16,4096,128,1,4178.913,68.035,2.254 -2026-07-16T06:13:04Z,29475951206,88,1,Vulkan Build and Benchmark,push,main,https://github.com/goniz/mlx-vulkan/actions/runs/29475951206,371c56655b502f49c021fe8ebea09d4f3ce7d686,430a3270612c8b0de2eec29a0f52c18c4fdc81b9,mlx-community/Qwen3-0.6B-8bit,8bit,4096,128,1,4137.078,119.812,1.997 -2026-07-16T06:13:04Z,29475951206,88,1,Vulkan Build and Benchmark,push,main,https://github.com/goniz/mlx-vulkan/actions/runs/29475951206,371c56655b502f49c021fe8ebea09d4f3ce7d686,430a3270612c8b0de2eec29a0f52c18c4fdc81b9,mlx-community/Qwen3.6-35B-A3B-8bit,8bit,4096,128,1,851.294,29.863,39.886 +timestamp_utc,github_run_id,github_run_number,github_run_attempt,github_workflow,ci_event,ci_ref,run_url,mlx_vulkan_commit,mlx_commit,model_name,model_bits,prompt_tokens,generation_tokens,batch_size,prompt_tps,generation_tps,peak_memory_gb +2026-05-18T08:58:24Z,,,,,legacy-readme,main,,5bdfd05f7804ce0a84e37c129330f9c377d13289,49f9a1df3917963752ba0cd299c8b9f430f4538a,mlx-community/Qwen3-0.6B-bf16,bf16,4096,128,1,2295.900,33.417,2.825 +2026-05-18T08:58:24Z,,,,,legacy-readme,main,,5bdfd05f7804ce0a84e37c129330f9c377d13289,49f9a1df3917963752ba0cd299c8b9f430f4538a,mlx-community/Qwen3-0.6B-8bit,8bit,4096,128,1,1317.353,20.398,2.823 +2026-05-18T11:10:19Z,,,,,local,,,5bdfd05f7804ce0a84e37c129330f9c377d13289,49f9a1df3917963752ba0cd299c8b9f430f4538a,mlx-community/Qwen3-0.6B-bf16,bf16,4096,128,1,2302.213,34.386,2.825 +2026-05-18T11:10:19Z,,,,,local,,,5bdfd05f7804ce0a84e37c129330f9c377d13289,49f9a1df3917963752ba0cd299c8b9f430f4538a,mlx-community/Qwen3-0.6B-8bit,8bit,4096,128,1,1324.214,20.660,2.823 +2026-05-18T11:40:12Z,26031178686,22,1,Vulkan Build and Benchmark,push,main,https://github.com/goniz/mlx-vulkan/actions/runs/26031178686,db88189558430aca509d9e15bb8a8f0a08ac83ef,49f9a1df3917963752ba0cd299c8b9f430f4538a,mlx-community/Qwen3-0.6B-bf16,bf16,4096,128,1,2282.178,33.745,2.825 +2026-05-18T11:40:12Z,26031178686,22,1,Vulkan Build and Benchmark,push,main,https://github.com/goniz/mlx-vulkan/actions/runs/26031178686,db88189558430aca509d9e15bb8a8f0a08ac83ef,49f9a1df3917963752ba0cd299c8b9f430f4538a,mlx-community/Qwen3-0.6B-8bit,8bit,4096,128,1,1302.198,18.662,2.823 +2026-05-19T17:57:06Z,26115334542,23,1,Vulkan Build and Benchmark,push,main,https://github.com/goniz/mlx-vulkan/actions/runs/26115334542,13821383e55fd9e88b693dd690565878e43a6bc9,f81649a8cfef762d21eb879149e663ce3dd8b794,mlx-community/Qwen3-0.6B-bf16,bf16,4096,128,1,2263.796,28.465,2.565 +2026-05-19T17:57:06Z,26115334542,23,1,Vulkan Build and Benchmark,push,main,https://github.com/goniz/mlx-vulkan/actions/runs/26115334542,13821383e55fd9e88b693dd690565878e43a6bc9,f81649a8cfef762d21eb879149e663ce3dd8b794,mlx-community/Qwen3-0.6B-8bit,8bit,4096,128,1,1316.035,19.709,1.917 +2026-05-20T05:17:15Z,26142834214,24,1,Vulkan Build and Benchmark,push,main,https://github.com/goniz/mlx-vulkan/actions/runs/26142834214,53a6764931412934649ade16840142594b933982,2a18b7c7a978764ec3cf606b28edc9f991dd4677,mlx-community/Qwen3-0.6B-bf16,bf16,4096,128,1,2337.462,28.049,2.565 +2026-05-20T05:17:15Z,26142834214,24,1,Vulkan Build and Benchmark,push,main,https://github.com/goniz/mlx-vulkan/actions/runs/26142834214,53a6764931412934649ade16840142594b933982,2a18b7c7a978764ec3cf606b28edc9f991dd4677,mlx-community/Qwen3-0.6B-8bit,8bit,4096,128,1,1330.239,19.969,1.917 +2026-05-20T09:39:03Z,26154299689,25,1,Vulkan Build and Benchmark,push,main,https://github.com/goniz/mlx-vulkan/actions/runs/26154299689,33eb2d2d013ebd151cf36c7ac841c70649cb5905,13c309699abde91be39f88b2b475fac9e3251412,mlx-community/Qwen3-0.6B-bf16,bf16,4096,128,1,2300.015,27.642,2.565 +2026-05-20T09:39:03Z,26154299689,25,1,Vulkan Build and Benchmark,push,main,https://github.com/goniz/mlx-vulkan/actions/runs/26154299689,33eb2d2d013ebd151cf36c7ac841c70649cb5905,13c309699abde91be39f88b2b475fac9e3251412,mlx-community/Qwen3-0.6B-8bit,8bit,4096,128,1,1266.470,17.653,1.917 +2026-05-20T10:17:07Z,26156115134,26,1,Vulkan Build and Benchmark,push,main,https://github.com/goniz/mlx-vulkan/actions/runs/26156115134,24c9e5a63ceaea177c13a9549ca117256663a6a6,37d3e21c98c25d9856ee12a2c0116b5df73bcfba,mlx-community/Qwen3-0.6B-bf16,bf16,4096,128,1,2270.406,28.463,2.565 +2026-05-20T10:17:07Z,26156115134,26,1,Vulkan Build and Benchmark,push,main,https://github.com/goniz/mlx-vulkan/actions/runs/26156115134,24c9e5a63ceaea177c13a9549ca117256663a6a6,37d3e21c98c25d9856ee12a2c0116b5df73bcfba,mlx-community/Qwen3-0.6B-8bit,8bit,4096,128,1,1313.222,18.601,1.917 +2026-05-20T17:03:01Z,26177458763,27,1,Vulkan Build and Benchmark,push,main,https://github.com/goniz/mlx-vulkan/actions/runs/26177458763,59b4efefffa5550badca78ceb86bfc7750e7f17b,7e13fc8fdb23c833129aed7f8a2ced8abf119f15,mlx-community/Qwen3-0.6B-bf16,bf16,4096,128,1,2289.089,28.488,2.565 +2026-05-20T17:03:01Z,26177458763,27,1,Vulkan Build and Benchmark,push,main,https://github.com/goniz/mlx-vulkan/actions/runs/26177458763,59b4efefffa5550badca78ceb86bfc7750e7f17b,7e13fc8fdb23c833129aed7f8a2ced8abf119f15,mlx-community/Qwen3-0.6B-8bit,8bit,4096,128,1,1317.869,18.364,1.919 +2026-05-20T17:44:14Z,26179663920,28,1,Vulkan Build and Benchmark,push,main,https://github.com/goniz/mlx-vulkan/actions/runs/26179663920,96fe5ce9a63b5b020bb66660a5718bf9a2cf6d84,e26ac9ffde2f6d9ca4638dd3c9ee2e45c67ede1b,mlx-community/Qwen3-0.6B-bf16,bf16,4096,128,1,2257.811,27.349,2.565 +2026-05-20T17:44:14Z,26179663920,28,1,Vulkan Build and Benchmark,push,main,https://github.com/goniz/mlx-vulkan/actions/runs/26179663920,96fe5ce9a63b5b020bb66660a5718bf9a2cf6d84,e26ac9ffde2f6d9ca4638dd3c9ee2e45c67ede1b,mlx-community/Qwen3-0.6B-8bit,8bit,4096,128,1,1304.978,18.946,1.917 +2026-05-24T18:08:49Z,26368847068,29,1,Vulkan Build and Benchmark,push,main,https://github.com/goniz/mlx-vulkan/actions/runs/26368847068,545e44c9fcb1e63f8776f8d05ebd44fed0a26fb3,045d17b6515c358c3e2cd9630216cb26995474ae,mlx-community/Qwen3-0.6B-bf16,bf16,4096,128,1,2286.532,28.557,2.603 +2026-05-24T18:08:49Z,26368847068,29,1,Vulkan Build and Benchmark,push,main,https://github.com/goniz/mlx-vulkan/actions/runs/26368847068,545e44c9fcb1e63f8776f8d05ebd44fed0a26fb3,045d17b6515c358c3e2cd9630216cb26995474ae,mlx-community/Qwen3-0.6B-8bit,8bit,4096,128,1,1320.984,18.960,1.896 +2026-05-24T19:50:46Z,26371082381,30,1,Vulkan Build and Benchmark,push,main,https://github.com/goniz/mlx-vulkan/actions/runs/26371082381,21e14fb3b4e190bad22184ccc3bd21f345e28344,68a4c6b9e3a4830286426fa645c081b327a23532,mlx-community/Qwen3-0.6B-bf16,bf16,4096,128,1,2264.269,28.568,2.603 +2026-05-24T19:50:46Z,26371082381,30,1,Vulkan Build and Benchmark,push,main,https://github.com/goniz/mlx-vulkan/actions/runs/26371082381,21e14fb3b4e190bad22184ccc3bd21f345e28344,68a4c6b9e3a4830286426fa645c081b327a23532,mlx-community/Qwen3-0.6B-8bit,8bit,4096,128,1,1311.879,18.768,1.901 +2026-05-24T20:02:13Z,,,,,local,,,21e14fb3b4e190bad22184ccc3bd21f345e28344,68a4c6b9e3a4830286426fa645c081b327a23532,mlx-community/Qwen3-0.6B-bf16,bf16,4096,128,1,2315.715,28.600,2.603 +2026-05-24T20:02:13Z,,,,,local,,,21e14fb3b4e190bad22184ccc3bd21f345e28344,68a4c6b9e3a4830286426fa645c081b327a23532,mlx-community/Qwen3-0.6B-8bit,8bit,4096,128,1,1328.602,19.271,1.901 +2026-05-24T20:21:37Z,26371767095,31,1,Vulkan Build and Benchmark,push,main,https://github.com/goniz/mlx-vulkan/actions/runs/26371767095,c1403315237c0e8c41d31c03fb4101d7632236d4,68a4c6b9e3a4830286426fa645c081b327a23532,mlx-community/Qwen3-0.6B-bf16,bf16,4096,128,1,2293.349,28.079,2.603 +2026-05-24T20:21:37Z,26371767095,31,1,Vulkan Build and Benchmark,push,main,https://github.com/goniz/mlx-vulkan/actions/runs/26371767095,c1403315237c0e8c41d31c03fb4101d7632236d4,68a4c6b9e3a4830286426fa645c081b327a23532,mlx-community/Qwen3-0.6B-8bit,8bit,4096,128,1,1320.912,18.808,1.898 +2026-05-24T22:57:10Z,26375042081,32,1,Vulkan Build and Benchmark,push,main,https://github.com/goniz/mlx-vulkan/actions/runs/26375042081,ea6f0f476b1ab75ca9c9657303d011b7fc90cff6,8e4a5d8200a0aa8acb02c401b291465b7f2deca8,mlx-community/Qwen3-0.6B-bf16,bf16,4096,128,1,2335.527,62.190,2.824 +2026-05-24T22:57:10Z,26375042081,32,1,Vulkan Build and Benchmark,push,main,https://github.com/goniz/mlx-vulkan/actions/runs/26375042081,ea6f0f476b1ab75ca9c9657303d011b7fc90cff6,8e4a5d8200a0aa8acb02c401b291465b7f2deca8,mlx-community/Qwen3-0.6B-8bit,8bit,4096,128,1,1329.371,25.035,2.204 +2026-05-25T05:38:25Z,26385119043,33,1,Vulkan Build and Benchmark,push,main,https://github.com/goniz/mlx-vulkan/actions/runs/26385119043,514902c9c32f1df2e98b631e79c458585092d0f8,51787f26d9dd78e4e88a117ca45256eb30f441bd,mlx-community/Qwen3-0.6B-bf16,bf16,4096,128,1,2348.431,67.708,2.824 +2026-05-25T05:38:25Z,26385119043,33,1,Vulkan Build and Benchmark,push,main,https://github.com/goniz/mlx-vulkan/actions/runs/26385119043,514902c9c32f1df2e98b631e79c458585092d0f8,51787f26d9dd78e4e88a117ca45256eb30f441bd,mlx-community/Qwen3-0.6B-8bit,8bit,4096,128,1,1326.742,26.004,2.204 +2026-05-25T16:38:02Z,26410543427,34,1,Vulkan Build and Benchmark,push,main,https://github.com/goniz/mlx-vulkan/actions/runs/26410543427,b71d53e623a47597a1412ca379c10247df88774d,aab59b9e2015cc83c891470d1f106ea3b7a831ed,mlx-community/Qwen3-0.6B-bf16,bf16,4096,128,1,2334.922,67.640,2.824 +2026-05-25T16:38:02Z,26410543427,34,1,Vulkan Build and Benchmark,push,main,https://github.com/goniz/mlx-vulkan/actions/runs/26410543427,b71d53e623a47597a1412ca379c10247df88774d,aab59b9e2015cc83c891470d1f106ea3b7a831ed,mlx-community/Qwen3-0.6B-8bit,8bit,4096,128,1,1302.833,26.039,2.204 +2026-05-27T07:30:15Z,26497290169,35,1,Vulkan Build and Benchmark,push,main,https://github.com/goniz/mlx-vulkan/actions/runs/26497290169,68a02370f0e50c00ac7bb8a630a3462286ed88af,ea869877c6948c826beb70ee4d7f5dac3739285f,mlx-community/Qwen3-0.6B-bf16,bf16,4096,128,1,2343.234,63.423,2.824 +2026-05-27T07:30:15Z,26497290169,35,1,Vulkan Build and Benchmark,push,main,https://github.com/goniz/mlx-vulkan/actions/runs/26497290169,68a02370f0e50c00ac7bb8a630a3462286ed88af,ea869877c6948c826beb70ee4d7f5dac3739285f,mlx-community/Qwen3-0.6B-8bit,8bit,4096,128,1,1331.303,28.681,2.204 +2026-05-28T09:45:35Z,26567238160,40,1,Vulkan Build and Benchmark,push,main,https://github.com/goniz/mlx-vulkan/actions/runs/26567238160,d922a89538259f01357550d2b1054415234d542a,38617fc360770cb7f8e448dd828c77f8eb860eab,mlx-community/Qwen3-0.6B-bf16,bf16,4096,128,1,2332.092,62.990,2.824 +2026-05-28T09:45:35Z,26567238160,40,1,Vulkan Build and Benchmark,push,main,https://github.com/goniz/mlx-vulkan/actions/runs/26567238160,d922a89538259f01357550d2b1054415234d542a,38617fc360770cb7f8e448dd828c77f8eb860eab,mlx-community/Qwen3-0.6B-8bit,8bit,4096,128,1,1329.738,27.392,2.242 +2026-05-28T14:42:43Z,26581733030,41,1,Vulkan Build and Benchmark,push,main,https://github.com/goniz/mlx-vulkan/actions/runs/26581733030,00dfd9c235a436d59bb3a327ba70bb455c221ae7,4a6c4a9e66278c18db0e4649423c1e4bee97a279,mlx-community/Qwen3-0.6B-bf16,bf16,4096,128,1,2349.782,63.834,2.824 +2026-05-28T14:42:43Z,26581733030,41,1,Vulkan Build and Benchmark,push,main,https://github.com/goniz/mlx-vulkan/actions/runs/26581733030,00dfd9c235a436d59bb3a327ba70bb455c221ae7,4a6c4a9e66278c18db0e4649423c1e4bee97a279,mlx-community/Qwen3-0.6B-8bit,8bit,4096,128,1,1336.072,27.618,2.204 +2026-05-28T17:33:59Z,26591235365,42,1,Vulkan Build and Benchmark,push,main,https://github.com/goniz/mlx-vulkan/actions/runs/26591235365,de335807c49cd93cd1ded812b48c144c1cbed71a,4a6c4a9e66278c18db0e4649423c1e4bee97a279,mlx-community/Qwen3-0.6B-bf16,bf16,4096,128,1,2346.036,63.854,2.824 +2026-05-28T17:33:59Z,26591235365,42,1,Vulkan Build and Benchmark,push,main,https://github.com/goniz/mlx-vulkan/actions/runs/26591235365,de335807c49cd93cd1ded812b48c144c1cbed71a,4a6c4a9e66278c18db0e4649423c1e4bee97a279,mlx-community/Qwen3-0.6B-8bit,8bit,4096,128,1,1336.223,25.825,2.246 +2026-05-28T17:33:59Z,26591235365,42,1,Vulkan Build and Benchmark,push,main,https://github.com/goniz/mlx-vulkan/actions/runs/26591235365,de335807c49cd93cd1ded812b48c144c1cbed71a,4a6c4a9e66278c18db0e4649423c1e4bee97a279,mlx-community/Qwen3.6-35B-A3B-8bit,8bit,4096,128,1,56.402,8.268,42.442 +2026-05-31T17:12:42Z,26719038679,46,1,Vulkan Build and Benchmark,push,main,https://github.com/goniz/mlx-vulkan/actions/runs/26719038679,8fbf5f9e3d6a2a2d0f16b3cfb086cfb7bbfe815d,0a10dbf9d61295ae2544c763634a590eb2197dd9,mlx-community/Qwen3-0.6B-bf16,bf16,4096,128,1,2392.984,60.641,2.824 +2026-05-31T17:12:42Z,26719038679,46,1,Vulkan Build and Benchmark,push,main,https://github.com/goniz/mlx-vulkan/actions/runs/26719038679,8fbf5f9e3d6a2a2d0f16b3cfb086cfb7bbfe815d,0a10dbf9d61295ae2544c763634a590eb2197dd9,mlx-community/Qwen3-0.6B-8bit,8bit,4096,128,1,1319.221,25.538,2.204 +2026-05-31T17:12:42Z,26719038679,46,1,Vulkan Build and Benchmark,push,main,https://github.com/goniz/mlx-vulkan/actions/runs/26719038679,8fbf5f9e3d6a2a2d0f16b3cfb086cfb7bbfe815d,0a10dbf9d61295ae2544c763634a590eb2197dd9,mlx-community/Qwen3.6-35B-A3B-8bit,8bit,4096,128,1,54.350,8.022,42.438 +2026-06-04T06:13:41Z,26934302484,49,1,Vulkan Build and Benchmark,push,main,https://github.com/goniz/mlx-vulkan/actions/runs/26934302484,23a25d1b8968fa3de6053b11b25cf4a69b97d45c,bc2fdc59c9a3326133aaf1ff6c607c9203ecf47f,mlx-community/Qwen3-0.6B-bf16,bf16,4096,128,1,2333.071,63.860,2.824 +2026-06-04T06:13:41Z,26934302484,49,1,Vulkan Build and Benchmark,push,main,https://github.com/goniz/mlx-vulkan/actions/runs/26934302484,23a25d1b8968fa3de6053b11b25cf4a69b97d45c,bc2fdc59c9a3326133aaf1ff6c607c9203ecf47f,mlx-community/Qwen3-0.6B-8bit,8bit,4096,128,1,1344.868,30.731,2.204 +2026-06-04T06:13:41Z,26934302484,49,1,Vulkan Build and Benchmark,push,main,https://github.com/goniz/mlx-vulkan/actions/runs/26934302484,23a25d1b8968fa3de6053b11b25cf4a69b97d45c,bc2fdc59c9a3326133aaf1ff6c607c9203ecf47f,mlx-community/Qwen3.6-35B-A3B-8bit,8bit,4096,128,1,57.111,8.362,42.440 +2026-06-04T07:22:00Z,26937061971,51,1,Vulkan Build and Benchmark,push,main,https://github.com/goniz/mlx-vulkan/actions/runs/26937061971,c0c3da7aa159b41cb29e469ba481a32843c2350f,5d618c8cecd05b0518721b5450d8e79ad9ab236d,mlx-community/Qwen3-0.6B-bf16,bf16,4096,128,1,2410.127,65.748,2.614 +2026-06-04T07:22:00Z,26937061971,51,1,Vulkan Build and Benchmark,push,main,https://github.com/goniz/mlx-vulkan/actions/runs/26937061971,c0c3da7aa159b41cb29e469ba481a32843c2350f,5d618c8cecd05b0518721b5450d8e79ad9ab236d,mlx-community/Qwen3-0.6B-8bit,8bit,4096,128,1,1360.581,87.609,2.056 +2026-06-04T07:22:00Z,26937061971,51,1,Vulkan Build and Benchmark,push,main,https://github.com/goniz/mlx-vulkan/actions/runs/26937061971,c0c3da7aa159b41cb29e469ba481a32843c2350f,5d618c8cecd05b0518721b5450d8e79ad9ab236d,mlx-community/Qwen3.6-35B-A3B-8bit,8bit,4096,128,1,121.011,21.135,40.350 +2026-06-13T22:02:47Z,27480356548,52,1,Vulkan Build and Benchmark,push,main,https://github.com/goniz/mlx-vulkan/actions/runs/27480356548,d1c89c732e398e2a469bbc206a5292f02f4c677a,75812951eedf0a7f82b1898e56c2281500d2e6c3,mlx-community/Qwen3-0.6B-bf16,bf16,4096,128,1,2860.579,65.949,2.614 +2026-06-13T22:02:47Z,27480356548,52,1,Vulkan Build and Benchmark,push,main,https://github.com/goniz/mlx-vulkan/actions/runs/27480356548,d1c89c732e398e2a469bbc206a5292f02f4c677a,75812951eedf0a7f82b1898e56c2281500d2e6c3,mlx-community/Qwen3-0.6B-8bit,8bit,4096,128,1,1477.926,87.524,2.056 +2026-06-13T22:02:47Z,27480356548,52,1,Vulkan Build and Benchmark,push,main,https://github.com/goniz/mlx-vulkan/actions/runs/27480356548,d1c89c732e398e2a469bbc206a5292f02f4c677a,75812951eedf0a7f82b1898e56c2281500d2e6c3,mlx-community/Qwen3.6-35B-A3B-8bit,8bit,4096,128,1,124.666,21.113,40.349 +2026-06-18T15:23:39Z,27769330530,57,1,Vulkan Build and Benchmark,push,main,https://github.com/goniz/mlx-vulkan/actions/runs/27769330530,4f515907bd714513c82e0426c219804b722a58eb,f799cc26d2f4703872d27b945d6c65ab528b4dee,mlx-community/Qwen3-0.6B-bf16,bf16,4096,128,1,2860.080,65.950,2.614 +2026-06-18T15:23:39Z,27769330530,57,1,Vulkan Build and Benchmark,push,main,https://github.com/goniz/mlx-vulkan/actions/runs/27769330530,4f515907bd714513c82e0426c219804b722a58eb,f799cc26d2f4703872d27b945d6c65ab528b4dee,mlx-community/Qwen3-0.6B-8bit,8bit,4096,128,1,1464.916,87.609,2.056 +2026-06-18T15:23:39Z,27769330530,57,1,Vulkan Build and Benchmark,push,main,https://github.com/goniz/mlx-vulkan/actions/runs/27769330530,4f515907bd714513c82e0426c219804b722a58eb,f799cc26d2f4703872d27b945d6c65ab528b4dee,mlx-community/Qwen3.6-35B-A3B-8bit,8bit,4096,128,1,120.009,21.288,40.350 +2026-06-19T15:37:13Z,,,,,local,,,baaf71e88ab18a1fe4abde2059b4eba48f4ddbb6,f799cc26d2f4703872d27b945d6c65ab528b4dee,mlx-community/Qwen3-0.6B-bf16,bf16,4096,128,1,2862.407,65.486,2.614 +2026-06-19T15:37:13Z,,,,,local,,,baaf71e88ab18a1fe4abde2059b4eba48f4ddbb6,f799cc26d2f4703872d27b945d6c65ab528b4dee,mlx-community/Qwen3-0.6B-8bit,8bit,4096,128,1,1458.232,85.857,2.055 +2026-06-19T15:37:13Z,,,,,local,,,baaf71e88ab18a1fe4abde2059b4eba48f4ddbb6,f799cc26d2f4703872d27b945d6c65ab528b4dee,mlx-community/Qwen3.6-35B-A3B-8bit,8bit,4096,128,1,120.704,20.996,40.349 +2026-07-02T08:24:34Z,28575944585,58,1,Vulkan Build and Benchmark,push,main,https://github.com/goniz/mlx-vulkan/actions/runs/28575944585,28e7e391e6e8a3d7f2afe15827d3fa7203d8a08c,088d06620ed01275ccbff7a16fda30348b563525,mlx-community/Qwen3-0.6B-bf16,bf16,4096,128,1,2818.647,65.027,2.614 +2026-07-02T08:24:34Z,28575944585,58,1,Vulkan Build and Benchmark,push,main,https://github.com/goniz/mlx-vulkan/actions/runs/28575944585,28e7e391e6e8a3d7f2afe15827d3fa7203d8a08c,088d06620ed01275ccbff7a16fda30348b563525,mlx-community/Qwen3-0.6B-8bit,8bit,4096,128,1,1470.973,85.274,2.056 +2026-07-02T08:24:34Z,28575944585,58,1,Vulkan Build and Benchmark,push,main,https://github.com/goniz/mlx-vulkan/actions/runs/28575944585,28e7e391e6e8a3d7f2afe15827d3fa7203d8a08c,088d06620ed01275ccbff7a16fda30348b563525,mlx-community/Qwen3.6-35B-A3B-8bit,8bit,4096,128,1,127.088,20.969,40.349 +2026-07-03T13:26:49Z,28663462997,63,1,Vulkan Build and Benchmark,push,main,https://github.com/goniz/mlx-vulkan/actions/runs/28663462997,ebd8aec5b30fb07ac40dca961277ef96821fe5e0,a00c15f18df4109f4b97a401dfeafe0ea7ce41f2,mlx-community/Qwen3-0.6B-bf16,bf16,4096,128,1,2841.434,62.497,2.723 +2026-07-03T13:26:49Z,28663462997,63,1,Vulkan Build and Benchmark,push,main,https://github.com/goniz/mlx-vulkan/actions/runs/28663462997,ebd8aec5b30fb07ac40dca961277ef96821fe5e0,a00c15f18df4109f4b97a401dfeafe0ea7ce41f2,mlx-community/Qwen3-0.6B-8bit,8bit,4096,128,1,1493.533,83.962,2.165 +2026-07-03T13:26:49Z,28663462997,63,1,Vulkan Build and Benchmark,push,main,https://github.com/goniz/mlx-vulkan/actions/runs/28663462997,ebd8aec5b30fb07ac40dca961277ef96821fe5e0,a00c15f18df4109f4b97a401dfeafe0ea7ce41f2,mlx-community/Qwen3.6-35B-A3B-8bit,8bit,4096,128,1,129.037,21.320,40.039 +2026-07-04T21:20:51Z,28719951837,65,1,Vulkan Build and Benchmark,push,main,https://github.com/goniz/mlx-vulkan/actions/runs/28719951837,f49be611be9644890d3491f70f693828187a7045,e639eb1189d304d6e0e4bc6b6caabc77640fe50e,mlx-community/Qwen3-0.6B-bf16,bf16,4096,128,1,2812.254,66.926,2.101 +2026-07-04T21:20:51Z,28719951837,65,1,Vulkan Build and Benchmark,push,main,https://github.com/goniz/mlx-vulkan/actions/runs/28719951837,f49be611be9644890d3491f70f693828187a7045,e639eb1189d304d6e0e4bc6b6caabc77640fe50e,mlx-community/Qwen3-0.6B-8bit,8bit,4096,128,1,1492.104,89.988,1.554 +2026-07-04T21:20:51Z,28719951837,65,1,Vulkan Build and Benchmark,push,main,https://github.com/goniz/mlx-vulkan/actions/runs/28719951837,f49be611be9644890d3491f70f693828187a7045,e639eb1189d304d6e0e4bc6b6caabc77640fe50e,mlx-community/Qwen3.6-35B-A3B-8bit,8bit,4096,128,1,126.534,21.898,39.235 +2026-07-07T07:47:39Z,28850130432,67,1,Vulkan Build and Benchmark,push,main,https://github.com/goniz/mlx-vulkan/actions/runs/28850130432,61e55afc60f7ba441fa3d1ba1fe4eba4d0384ccf,d2f75a158dd8d16c7f5dd7aced3d598d38767b46,mlx-community/Qwen3-0.6B-bf16,bf16,4096,128,1,2833.533,66.528,2.102 +2026-07-07T07:47:39Z,28850130432,67,1,Vulkan Build and Benchmark,push,main,https://github.com/goniz/mlx-vulkan/actions/runs/28850130432,61e55afc60f7ba441fa3d1ba1fe4eba4d0384ccf,d2f75a158dd8d16c7f5dd7aced3d598d38767b46,mlx-community/Qwen3-0.6B-8bit,8bit,4096,128,1,1493.524,89.920,1.544 +2026-07-07T07:47:39Z,28850130432,67,1,Vulkan Build and Benchmark,push,main,https://github.com/goniz/mlx-vulkan/actions/runs/28850130432,61e55afc60f7ba441fa3d1ba1fe4eba4d0384ccf,d2f75a158dd8d16c7f5dd7aced3d598d38767b46,mlx-community/Qwen3.6-35B-A3B-8bit,8bit,4096,128,1,125.717,21.956,39.235 +2026-07-09T11:58:18Z,29016445983,69,1,Vulkan Build and Benchmark,push,main,https://github.com/goniz/mlx-vulkan/actions/runs/29016445983,8cb61017e9bde777026fde6c4831210fca32ec82,c7c199f79df4b987c134860a92430d660cc312ad,mlx-community/Qwen3-0.6B-bf16,bf16,4096,128,1,2825.530,66.992,2.099 +2026-07-09T11:58:18Z,29016445983,69,1,Vulkan Build and Benchmark,push,main,https://github.com/goniz/mlx-vulkan/actions/runs/29016445983,8cb61017e9bde777026fde6c4831210fca32ec82,c7c199f79df4b987c134860a92430d660cc312ad,mlx-community/Qwen3-0.6B-8bit,8bit,4096,128,1,1493.076,90.933,1.547 +2026-07-09T11:58:18Z,29016445983,69,1,Vulkan Build and Benchmark,push,main,https://github.com/goniz/mlx-vulkan/actions/runs/29016445983,8cb61017e9bde777026fde6c4831210fca32ec82,c7c199f79df4b987c134860a92430d660cc312ad,mlx-community/Qwen3.6-35B-A3B-8bit,8bit,4096,128,1,127.692,22.127,39.235 +2026-07-09T12:42:46Z,29018390272,71,1,Vulkan Build and Benchmark,push,main,https://github.com/goniz/mlx-vulkan/actions/runs/29018390272,6c1d1aa78607ad8c47df6fdc3eb4357f4b319380,4c86a8302d94e33417d45faaef9d551e05d8981b,mlx-community/Qwen3-0.6B-bf16,bf16,4096,128,1,2819.923,66.982,2.103 +2026-07-09T12:42:46Z,29018390272,71,1,Vulkan Build and Benchmark,push,main,https://github.com/goniz/mlx-vulkan/actions/runs/29018390272,6c1d1aa78607ad8c47df6fdc3eb4357f4b319380,4c86a8302d94e33417d45faaef9d551e05d8981b,mlx-community/Qwen3-0.6B-8bit,8bit,4096,128,1,1497.589,89.741,1.552 +2026-07-09T12:42:46Z,29018390272,71,1,Vulkan Build and Benchmark,push,main,https://github.com/goniz/mlx-vulkan/actions/runs/29018390272,6c1d1aa78607ad8c47df6fdc3eb4357f4b319380,4c86a8302d94e33417d45faaef9d551e05d8981b,mlx-community/Qwen3.6-35B-A3B-8bit,8bit,4096,128,1,127.663,22.189,39.283 +2026-07-13T20:09:53Z,29281094972,72,1,Vulkan Build and Benchmark,push,main,https://github.com/goniz/mlx-vulkan/actions/runs/29281094972,fd80e5ae3b752e999fd92588f1acc7f5426ae0ee,277387e9966afe5774da27da5caa9e818a05fb94,mlx-community/Qwen3-0.6B-bf16,bf16,4096,128,1,2839.983,66.848,2.102 +2026-07-13T20:09:53Z,29281094972,72,1,Vulkan Build and Benchmark,push,main,https://github.com/goniz/mlx-vulkan/actions/runs/29281094972,fd80e5ae3b752e999fd92588f1acc7f5426ae0ee,277387e9966afe5774da27da5caa9e818a05fb94,mlx-community/Qwen3-0.6B-8bit,8bit,4096,128,1,2788.160,90.037,1.653 +2026-07-13T20:09:53Z,29281094972,72,1,Vulkan Build and Benchmark,push,main,https://github.com/goniz/mlx-vulkan/actions/runs/29281094972,fd80e5ae3b752e999fd92588f1acc7f5426ae0ee,277387e9966afe5774da27da5caa9e818a05fb94,mlx-community/Qwen3.6-35B-A3B-8bit,8bit,4096,128,1,148.147,22.187,39.245 +2026-07-14T14:36:20Z,29341360981,73,1,Vulkan Build and Benchmark,push,main,https://github.com/goniz/mlx-vulkan/actions/runs/29341360981,58d5f072e9d0dc42aafeac665f299ab685ebbc9a,501c4b5571a1e1be85f11e244473a67123b50465,mlx-community/Qwen3-0.6B-bf16,bf16,4096,128,1,2812.893,67.037,2.099 +2026-07-14T14:36:20Z,29341360981,73,1,Vulkan Build and Benchmark,push,main,https://github.com/goniz/mlx-vulkan/actions/runs/29341360981,58d5f072e9d0dc42aafeac665f299ab685ebbc9a,501c4b5571a1e1be85f11e244473a67123b50465,mlx-community/Qwen3-0.6B-8bit,8bit,4096,128,1,2820.932,90.042,1.671 +2026-07-14T14:36:20Z,29341360981,73,1,Vulkan Build and Benchmark,push,main,https://github.com/goniz/mlx-vulkan/actions/runs/29341360981,58d5f072e9d0dc42aafeac665f299ab685ebbc9a,501c4b5571a1e1be85f11e244473a67123b50465,mlx-community/Qwen3.6-35B-A3B-8bit,8bit,4096,128,1,526.499,21.922,39.260 +2026-07-15T10:47:58Z,29409233711,78,1,Vulkan Build and Benchmark,push,main,https://github.com/goniz/mlx-vulkan/actions/runs/29409233711,76e10ae33b93eec085bc509dd7d8b30be4420af0,df10f7b8fa535bf4b07e6ff6e3082f38565bd20f,mlx-community/Qwen3-0.6B-bf16,bf16,4096,128,1,2949.229,67.981,2.102 +2026-07-15T10:47:58Z,29409233711,78,1,Vulkan Build and Benchmark,push,main,https://github.com/goniz/mlx-vulkan/actions/runs/29409233711,76e10ae33b93eec085bc509dd7d8b30be4420af0,df10f7b8fa535bf4b07e6ff6e3082f38565bd20f,mlx-community/Qwen3-0.6B-8bit,8bit,4096,128,1,2882.832,121.007,1.688 +2026-07-15T10:47:58Z,29409233711,78,1,Vulkan Build and Benchmark,push,main,https://github.com/goniz/mlx-vulkan/actions/runs/29409233711,76e10ae33b93eec085bc509dd7d8b30be4420af0,df10f7b8fa535bf4b07e6ff6e3082f38565bd20f,mlx-community/Qwen3.6-35B-A3B-8bit,8bit,4096,128,1,683.012,29.409,39.665 +2026-07-16T06:13:04Z,29475951206,88,1,Vulkan Build and Benchmark,push,main,https://github.com/goniz/mlx-vulkan/actions/runs/29475951206,371c56655b502f49c021fe8ebea09d4f3ce7d686,430a3270612c8b0de2eec29a0f52c18c4fdc81b9,mlx-community/Qwen3-0.6B-bf16,bf16,4096,128,1,4178.913,68.035,2.254 +2026-07-16T06:13:04Z,29475951206,88,1,Vulkan Build and Benchmark,push,main,https://github.com/goniz/mlx-vulkan/actions/runs/29475951206,371c56655b502f49c021fe8ebea09d4f3ce7d686,430a3270612c8b0de2eec29a0f52c18c4fdc81b9,mlx-community/Qwen3-0.6B-8bit,8bit,4096,128,1,4137.078,119.812,1.997 +2026-07-16T06:13:04Z,29475951206,88,1,Vulkan Build and Benchmark,push,main,https://github.com/goniz/mlx-vulkan/actions/runs/29475951206,371c56655b502f49c021fe8ebea09d4f3ce7d686,430a3270612c8b0de2eec29a0f52c18c4fdc81b9,mlx-community/Qwen3.6-35B-A3B-8bit,8bit,4096,128,1,851.294,29.863,39.886 +2026-07-16T11:53:40Z,29495909657,89,1,Vulkan Build and Benchmark,workflow_dispatch,cursor/gdn-vulkan-dv-pack-a20e,https://github.com/goniz/mlx-vulkan/actions/runs/29495909657,e6b2da3a32e7e2f3c7fc4639f8a969208832d053,ea5289c915a6f433621037b2c8a90c10e01b8b2a,mlx-community/Qwen3-0.6B-bf16,bf16,4096,128,1,4178.176,67.994,2.262 +2026-07-16T11:53:40Z,29495909657,89,1,Vulkan Build and Benchmark,workflow_dispatch,cursor/gdn-vulkan-dv-pack-a20e,https://github.com/goniz/mlx-vulkan/actions/runs/29495909657,e6b2da3a32e7e2f3c7fc4639f8a969208832d053,ea5289c915a6f433621037b2c8a90c10e01b8b2a,mlx-community/Qwen3-0.6B-8bit,8bit,4096,128,1,4124.938,119.797,1.998 +2026-07-16T11:53:40Z,29495909657,89,1,Vulkan Build and Benchmark,workflow_dispatch,cursor/gdn-vulkan-dv-pack-a20e,https://github.com/goniz/mlx-vulkan/actions/runs/29495909657,e6b2da3a32e7e2f3c7fc4639f8a969208832d053,ea5289c915a6f433621037b2c8a90c10e01b8b2a,mlx-community/Qwen3.6-35B-A3B-8bit,8bit,4096,128,1,903.405,29.568,39.885 diff --git a/mlx b/mlx index 430a327..ea5289c 160000 --- a/mlx +++ b/mlx @@ -1 +1 @@ -Subproject commit 430a3270612c8b0de2eec29a0f52c18c4fdc81b9 +Subproject commit ea5289c915a6f433621037b2c8a90c10e01b8b2a diff --git a/mlx-lm b/mlx-lm index 5fd426d..384030c 160000 --- a/mlx-lm +++ b/mlx-lm @@ -1 +1 @@ -Subproject commit 5fd426dee2a4c9853b9569c95a97dd1a5053c0e3 +Subproject commit 384030c511b9eb91ddd24ebea6abe60c1151dc33