qwen38

Optimized SGLang runtime for Qwen3.8-27B FP8 with DFlash2 and Qwen3.8 Flash-Next NVFP4 with FR-Spec on one NVIDIA RTX PRO 6000 Blackwell 96 GB GPU (SM120): 524K context, HiCache and NIXL.

Qwen3.8-Flash-Next on 2× RTX 3090: with 128 GB RAM up to 4,191 tok/s prefill · 111.5 tok/s decode (131K prompt), full 256K window at 2,865 tok/s; with 64 GB RAM 3,410 tok/s prefill · 84 tok/s decode. New: opt-in uncensored mode (runtime abliteration, no new weights).

Qwen3.8-Flash-Next 177B MoE (ISTA GSQ-RCO IQ3_XXS) self-hosted on RTX 5090 Laptop (24GB VRAM + 64GB RAM) via Strata: up to 110 tok/s (100+ sustained), 256K full context, vision, MTP speculative decoding, reasoning default xhigh (max tier). GPU+CPU both saturated. 26 measured rounds, 7 quant tiers screened. 中英双语实测实录

Qwen3.8-Flash-Next on 4x RTX 3090 with vLLM: 924,993 kv pool, 160-175tok/s single decode, 3 x 262K sessions resident, MTP, host-mapped PLE, pinned build and container recipes.

Run Qwen3.8-Flash-Next on ONE RTX 3090 (24 GB) + 64 GB RAM: 128K context, up to 2,100 tok/s prefill, 43–51 tok/s decode. vLLM runtime with hot MoE experts on the GPU and cold experts computed on the CPU, INT8 KV cache, Docker, OpenAI-compatible API.

Qwen3.8-27B and the ThinkingCap fine-tune at 262K context on dual or quad RTX 3090s: Merlin engine v2.5.2 + 11 patches, FP8 KV pool 826,689 (dual) / 1,974,072 (quad), MTP K=4

Swift 1.5 post-train of Qwen3.8-Flash-Next on 4x RTX 3090 with vLLM: W4A16 checkpoint on the unchanged Flash-Next v2.5.1 engine, TP2 x PP2 + EP, 924,993-token FP8 KV pool, 262K context, MTP K=3, recalibrated KV scales.

Inference engines for Gemma, Qwen and Pocket TTS in pure Go. No cgo, no Python. One static binary, CPU or Vulkan.

Qwen3.8-Flash-Next on 2× RTX 3090: with 128 GB RAM up to 4,191 tok/s prefill · 111.5 tok/s decode (131K prompt), full 256K window at 2,865 tok/s; with 64 GB RAM 3,410 tok/s prefill · 84 tok/s decode. New: opt-in uncensored mode (runtime abliteration, no new weights).

A tower defense made in Three js to test Qwen3.8-27B performance ( hf download hf://unsloth/Qwen3.8-27B-GGUF/Qwen3.8-27B-UD-Q3_K_XL.gguf , 128K context )

MTP (Model-Translation-Postprocess) is a ground-breaking technology designed to enhance the capabilities of Qwen, a sophisticated language model

Qwen3

Local deployment of Ternary Bonsai 2 27B (PrismML 1.75-bit ternary quantized Qwen3.8 27B) on an RTX 4060 laptop — setup, benchmarks and notes

In the realm of artificial intelligence, the Qwen3

What is Qwen3

Qwen3

Qwen 3.8 builds The Backrooms

Qwen3

Qwen3.8-Flash-Next on 2× RTX 3090: with 128 GB RAM up to 4,191 tok/s prefill · 111.5 tok/s decode (131K prompt), full 256K window at 2,865 tok/s; with 64 GB RAM 3,410 tok/s prefill · 84 tok/s decode. New: opt-in uncensored mode (runtime abliteration, no new weights).