diff --git a/snapshots/anthropic/deprecations.md b/snapshots/anthropic/deprecations.md index 3fb366e..921d3da 100644 --- a/snapshots/anthropic/deprecations.md +++ b/snapshots/anthropic/deprecations.md @@ -65,13 +65,14 @@ At some point, Anthropic hopes to make past models publicly available again. In ## Model status - [Claude Mythos Preview](https://anthropic.com/glasswing) (`claude-mythos-preview`) is deprecated. To migrate to [Claude Mythos 5](https://anthropic.com/glasswing) (`claude-mythos-5`), see the [migration guide](https://platform.claude.com/docs/en/about-claude/models/migration-guide#migrating-from-claude-mythos-preview). + [Claude Mythos Preview](https://anthropic.com/glasswing) (`claude-mythos-preview`) is deprecated. To migrate to [Claude Mythos 5](https://anthropic.com/glasswing) (`claude-mythos-5`), see the [migration guide](https://platform.claude.com/docs/en/models/fable-5/migration-guide#migrating-from-claude-mythos-preview). Current and recently retired models are listed in the following table with their status: | API model name | Current state | Deprecated | Tentative retirement date | | -------------------------- | ------------- | ----------------- | ---------------------------------- | +| claude-fable-5-1 | Active | N/A | Not sooner than September 1, 2027 | | claude-fable-5 | Active | N/A | Not sooner than June 9, 2027 | | claude-opus-5 | Active | N/A | Not sooner than July 24, 2027 | | claude-opus-4-8 | Active | N/A | Not sooner than May 28, 2027 | diff --git a/snapshots/cerebras/deprecations.md b/snapshots/cerebras/deprecations.md index 9bd4e42..a6d9708 100644 --- a/snapshots/cerebras/deprecations.md +++ b/snapshots/cerebras/deprecations.md @@ -6,6 +6,14 @@ > A list of all deprecations, with the most recent announcements appearing first. + + **Gemma 4 31B availability changes on public endpoints** + + Starting September 3, 2026, `gemma-4-31b` is no longer available on Cerebras public endpoints. This change doesn't affect [Dedicated Endpoints](/dedicated/overview), where Gemma 4 31B remains available. + + Use [`qwen-3.8-27b`](/models/qwen-3.8-27b) for public endpoint workloads. To continue using Gemma 4 31B, [contact us](https://www.cerebras.ai/contact) to set up a dedicated endpoint. + + **Deprecated `zai-glm-4.7`** diff --git a/snapshots/cohere/models.md b/snapshots/cohere/models.md index fe85d22..363f5b0 100644 --- a/snapshots/cohere/models.md +++ b/snapshots/cohere/models.md @@ -34,8 +34,9 @@ are. [retrieval augmented generation](retrieval-augmented-generation-rag) (RAG), translation, copywriting, and similar use cases. They work through the [Chat](../reference/chat) endpoint, which can be used with or without RAG. -* [Rerank](https://cohere.com/blog/rerank/?_gl=1*1t6ls4x*_ga*MTAxNTg1NTM1MS4xNjk1MjMwODQw*_ga_CRGS116RZS*MTcxNzYwMzYxMy4zNTEuMS4xNzE3NjAzNjUxLjIyLjAuMA..) is the fastest way to inject the intelligence of a language model into an existing search system. It can be accessed via the [Rerank](../reference/rerank-1) endpoint. -* [Embed](https://cohere.com/models/embed?_gl=1*1t6ls4x*_ga*MTAxNTg1NTM1MS4xNjk1MjMwODQw*_ga_CRGS116RZS*MTcxNzYwMzYxMy4zNTEuMS4xNzE3NjAzNjUxLjIyLjAuMA..) improves the accuracy of search, classification, clustering, and RAG results. It powers the [Embed](../reference/embed) endpoint. +* [Rerank](../../docs/rerank) is the fastest way to inject the intelligence of a language model into an existing search system. It can be accessed via the [Rerank](../reference/rerank-1) endpoint. +* [Embed](../../docs/cohere-embed) improves the accuracy of search, classification, clustering, and RAG results. It powers the [Embed](../reference/embed) endpoint. +* [Parse](../../docs/parse) extracts structured, machine-readable data from unstructured enterprise documents, like forms and PDFs. It powers the [Parse](../../v2/reference/parse) endpoint. * [Cohere Transcribe](../../v2/docs/transcribe) is Cohere's dedicated audio transcription model for automatic speech recognition (ASR). It powers the [Audio Transcriptions](../../v2/reference/create-audio-transcription) endpoint. [Cohere Transcribe Arabic](../../v2/docs/transcribe-arabic) is a version of the model optimized for Arabic-language @@ -88,13 +89,13 @@ In this table, we provide some important context for using Cohere Command models These models can be used to generate embeddings from text or classify it based on various parameters. Embeddings can be used for estimating semantic similarity between two sentences, choosing a sentence which is most likely to follow another sentence, or categorizing user feedback. The Representation model comes with a variety of helper functions, such as for detecting the language of an input. -| Model Name | Description | Modalities | Dimensions | Context Length | Similarity Metric | Endpoints | -| ------------------------------- | ------------------------------------------------------------------------------------------------------------------------- | -------------------------------------------- | ------------------------------------------ | -------------- | ------------------------------------------------------------- | ------------------------------------------------------------------------- | -| `embed-v4.0` | A model that allows for text and images to be classified or turned into embeddings | Text, Images, Mixed texts/images (i.e. PDFs) | One of '\[256, 512, 1024, 1536 (default)]' | 128k | Cosine Similarity, Dot Product Similarity, Euclidean Distance | [Embed](../reference/embed),
[Embed Jobs](../reference/embed-jobs) | -| `embed-english-v3.0` | A model that allows for text to be classified or turned into embeddings. English only. | Text, Images | 1024 | 512 | Cosine Similarity | [Embed](../reference/embed),
[Embed Jobs](../reference/embed-jobs) | -| `embed-english-light-v3.0` | A smaller, faster version of `embed-english-v3.0`. Almost as capable, but a lot faster. English only. | Text, Images | 384 | 512 | Cosine Similarity | [Embed](../reference/embed),
[Embed Jobs](../reference/embed-jobs) | -| `embed-multilingual-v3.0` | Provides multilingual classification and embedding support. [See supported languages here.](/docs/supported-languages) | Text, Images | 1024 | 512 | Cosine Similarity | [Embed](../reference/embed), [Embed Jobs](../reference/embed-jobs) | -| `embed-multilingual-light-v3.0` | A smaller, faster version of `embed-multilingual-v3.0`. Almost as capable, but a lot faster. Supports multiple languages. | Text, Images | 384 | 512 | Cosine Similarity | [Embed](../reference/embed),
[Embed Jobs](../reference/embed-jobs) | +| Model Name | Description | Modalities | Dimensions | Context Length | Similarity Metric | Endpoints | +| ------------------------------- | ------------------------------------------------------------------------------------------------------------------------- | -------------------------------------------- | ------------------------------------------ | -------------- | ------------------------------------------------------------- | ------------------------------------------------------------------- | +| `embed-v4.0` | A model that allows for text and images to be classified or turned into embeddings | Text, Images, Mixed texts/images (i.e. PDFs) | One of '\[256, 512, 1024, 1536 (default)]' | 128k | Cosine Similarity, Dot Product Similarity, Euclidean Distance | [Embed](../reference/embed), [Embed Jobs](../reference/embed-jobs) | +| `embed-english-v3.0` | A model that allows for text to be classified or turned into embeddings. English only. | Text, Images | 1024 | 512 | Cosine Similarity | [Embed](../reference/embed), [Embed Jobs](../reference/embed-jobs) | +| `embed-english-light-v3.0` | A smaller, faster version of `embed-english-v3.0`. Almost as capable, but a lot faster. English only. | Text, Images | 384 | 512 | Cosine Similarity | [Embed](../reference/embed), [Embed Jobs](../reference/embed-jobs) | +| `embed-multilingual-v3.0` | Provides multilingual classification and embedding support. [See supported languages here.](/docs/supported-languages) | Text, Images | 1024 | 512 | Cosine Similarity | [Embed](../reference/embed), [Embed Jobs](../reference/embed-jobs) | +| `embed-multilingual-light-v3.0` | A smaller, faster version of `embed-multilingual-v3.0`. Almost as capable, but a lot faster. Supports multiple languages. | Text, Images | 384 | 512 | Cosine Similarity | [Embed](../reference/embed), [Embed Jobs](../reference/embed-jobs) | ### Using Embed Models on Different Platforms @@ -135,17 +136,35 @@ In this table, we provide some important context for using Cohere Rerank models | `rerank-english-v3.0` | N/A | Unique per deployment | `Cohere-rerank-v3-english` | N/A | | `rerank-multilingual-v3.0` | N/A | Unique per deployment | `Cohere-rerank-v3-multilingual` | N/A | -
+\ + Rerank accepts full strings rather than tokens, so the token limit works a little differently. Rerank will automatically chunk documents longer than 510 tokens, and there is therefore no explicit limit to how long a document can be when using rerank. See our [best practice guide](/docs/reranking-best-practices) for more info about formatting documents for the Rerank endpoint. +## Parse + +Parse is our vision parsing model for extracting structured data from enterprise documents that can be used in AI search and agentic applications. + +| Model Name | Status | Description | Endpoints | +| ------------ | ------ | ----------------------------------------------------------------------------- | --------------------------------- | +| `parse-v5.0` | Live | For document intelligence workloads that demand accurate, high-volume parsing | [Parse](../../v2/reference/parse) | + +### Using Parse Models on Different Platforms + +In this table, we provide some important context for using Cohere Parse on Amazon SageMaker and Azure Foundry. + +| Model Name | Amazon SageMaker | Azure AI Foundry | +| :----------- | :-------------------- | :------------------ | +| `parse-v5.0` | Unique per deployment | `cohere-parse-v5.0` | + ## Audio Cohere Transcribe is our dedicated model for audio-in, text-out automatic speech recognition (ASR) workloads. For Arabic-language transcription, use Cohere Transcribe Arabic for best-in-class performance. -| Model Name | Status | Description | Maximum file size | Endpoints | -| --------------------------- | ------ | ----------------------------------------------------------------------------------------- | ----------------- | --------------------------------------------------------------------- | -| `cohere-transcribe-03-2026` | Live | Open source research release focused on high-accuracy, multilingual speech transcription. | 25MB | [Audio Transcriptions](../../v2/reference/create-audio-transcription) | +| Model Name | Status | Description | Maximum file size | Endpoints | +| ---------------------------------- | ------ | ------------------------------------------------------------------------------ | ----------------- | --------------------------------------------------------------------- | +| `cohere-transcribe-03-2026` | Live | Open source model focused on high-accuracy, multilingual speech transcription. | 25MB | [Audio Transcriptions](../../v2/reference/create-audio-transcription) | +| `cohere-transcribe-arabic-07-2026` | Live | Finetune optimized for Arabic audio inputs | 25MB | [Audio Transcriptions](../../v2/reference/create-audio-transcription) | ### Using Audio Models on Different Platforms diff --git a/snapshots/fireworks/changelog.md b/snapshots/fireworks/changelog.md index 0759d77..0734ade 100644 --- a/snapshots/fireworks/changelog.md +++ b/snapshots/fireworks/changelog.md @@ -4,19 +4,99 @@ # Changelog + + Platform + + # Deployment tags and annotation API changes + + Deployment tags are customer-managed entries stored in a deployment's `annotations` map. firectl presents logical keys such as `environment`; the REST API represents the same key as `custom/environment`. + + * **firectl:** Version 1.8.3 adds `deployment tag set`, `unset`, and `list`, including atomic batch operations. + * **REST writes:** Customer-managed annotation keys must begin with `custom/`. Bare keys now return HTTP `403` (`PERMISSION_DENIED`). + * **REST reads:** For regular account users, `GetDeployment` and `ListDeployments` return only `custom/*` annotation entries. Keys outside that namespace are omitted without an error. + + Existing stored annotations were not rewritten. Clients using a bare key such as `environment` should set `custom/environment` and update reads to use that canonical key. + + See [Deployment Tags](/deployments/deployment-tags) for commands, REST examples, validation rules, and migration guidance. + + + + Inference + + # Serverless rate limit ceilings now scale by model size + + Serverless adaptive rate limit ceilings now vary by model size tier. Smaller models (\< 400B parameters) get higher ceilings; medium models (400B – \< 1.6T) get intermediate ceilings; large models (≥ 1.6T) keep the previous base ceilings. See [Serverless rate limits](/serverless/rate-limits#model-size-tiers) for tier thresholds and ceiling values. + + + + Training + + # New Serverless Training models: DeepSeek V4 Flash 0731, Qwen 3.8 27B, and Muse Glimmer 30B + + The following models are now available for LoRA workloads on the shared Serverless Training pool: + + * **[DeepSeek V4 Flash 0731](https://app.fireworks.ai/models/fireworks/deepseek-v4-flash-0731)** with up to 262K context + * **[Qwen 3.8 27B](https://app.fireworks.ai/models/fireworks/qwen3p8-27b)** with up to 128K context + * **[Muse Glimmer 30B](https://app.fireworks.ai/models/fireworks/muse-glimmer-30b)** with up to 128K context + + See the [Serverless Training guide](/fine-tuning/training-api/serverless) for setup and the [training model catalog](/fine-tuning/models) for current availability. + + + + Inference + + # Serverless deprecation: MiniMax M2.7, GPT OSS 20B, Kimi K2.6 Turbo/Fast, Kimi K2.7 Code Fast, DeepSeek V4 Pro + + The following models are deprecated from serverless effective August 27, 2026. + + ## **Recommended migrations** + + * **[MiniMax M2.7](https://app.fireworks.ai/models/fireworks/minimax-m2p7)** — migrate to **[MiniMax M3](https://app.fireworks.ai/models/fireworks/minimax-m3)** + * **[GPT OSS 20B](https://app.fireworks.ai/models/fireworks/gpt-oss-20b)** — migrate to **[GPT OSS 120B](https://app.fireworks.ai/models/fireworks/gpt-oss-120b)** or **[Qwen3 8B](https://app.fireworks.ai/models/fireworks/qwen3-8b)** for lower-latency workloads + * **[Kimi K2.6 Turbo / Fast](https://app.fireworks.ai/models/fireworks/kimi-k2p6)** — migrate to **[Kimi K2.6](https://app.fireworks.ai/models/fireworks/kimi-k2p6)** (standard serving path) + * **[Kimi K2.7 Code Fast](https://app.fireworks.ai/models/fireworks/kimi-k2p7-code)** — migrate to **[Kimi K2.7 Code](https://app.fireworks.ai/models/fireworks/kimi-k2p7-code)** (standard serving path) + * **[DeepSeek V4 Pro](https://app.fireworks.ai/models/fireworks/deepseek-v4-pro)** — migrate to **[DeepSeek V4 Pro (0813)](https://app.fireworks.ai/models/fireworks/deepseek-v4-pro-0813)** + + + + Training + + # Serverless Training deprecation: Qwen 3.5 9B and Qwen 3.6 27B + + **Qwen 3.5 9B** and **Qwen 3.6 27B** are deprecated from Serverless Training effective August 26, 2026. + + Migrate new and existing Serverless Training workloads to **[Qwen 3.8 27B](https://app.fireworks.ai/models/fireworks/qwen3p8-27b)**. + + This change applies to the shared Serverless Training pool. Check the [training model catalog](/fine-tuning/models) for availability on other training surfaces. + + + + Platform + + # SSO documentation: IdP-initiated SAML + + Updated the [Custom SSO](/accounts/sso) guide. IdP-initiated SAML is supported as an opt-in (`--enable-idp-initiated-sso`); the previous troubleshooting copy that said Fireworks only supported SP-initiated login was incorrect. + + + Inference + # Serverless deprecation: DeepSeek V4 Flash **DeepSeek V4 Flash** is deprecated from serverless. Migrate to **[DeepSeek V4 Flash (0731)](https://app.fireworks.ai/models/fireworks/deepseek-v4-flash-0731)**. + Training + # SFT learning rate scheduler documentation Documented learning rate scheduler settings for supervised fine-tuning jobs, including `constant`, `linear`, and `cosine` schedules via `firectl` and the REST API `lrScheduler` object. + Inference + # Serverless deprecation: Kimi K2.5 and Qwen 3.6 Plus **Kimi K2.5** and **Qwen 3.6 Plus** are deprecated from serverless. @@ -28,18 +108,24 @@ + Inference + # Serverless deprecation: MiniMax M2.5 **MiniMax M2.5** is deprecated from serverless. Migrate to **[MiniMax M2.7](https://app.fireworks.ai/models/fireworks/minimax-m2p7)**. + Inference + # New model: GLM 5.2 **[GLM 5.2](https://app.fireworks.ai/models/fireworks/glm-5p2)** is now available in the Model Library. + Inference + # New models: Kimi K2.7 Code, MiniMax M3, and Qwen 3.7 Plus The following models are now available in the Model Library: @@ -50,12 +136,16 @@ + Inference + # Audio inference and image generation deprecation Audio inference and image generation are deprecated. + Inference + # Serverless deprecation: legacy models removed May 14, 2026 Several legacy serverless models will be decommissioned on **May 14, 2026** to make room for newer, higher-performance releases. This applies **only to serverless** usage; **dedicated deployments are unaffected**. @@ -79,6 +169,8 @@ + Inference Training Platform + # Video & Audio Models, AWS S3 Training Integration, and SSO Improvements ## **Video & Audio Input Models** @@ -91,7 +183,7 @@ Training datasets can now be stored in your own AWS S3 buckets using GCP-to-AWS OIDC federation. This Bring Your Own Bucket (BYOB) approach keeps your data private while enabling secure access during Supervised Fine-Tuning and Reinforcement Fine-Tuning jobs—no long-lived credentials required. - See the [Secure Training (BYOB) documentation](/fine-tuning/finetuning-intro#training-security) for IAM role setup and usage examples. + See the [Secure Training (BYOB) documentation](/guides/security_compliance/secure_training) for IAM role setup and usage examples. ## **Just-In-Time (JIT) User Provisioning for SSO** (Enterprise) @@ -102,7 +194,7 @@ ## 📚 Documentation Updates * **Video & Audio Inputs:** New guide for processing video and audio with Qwen3 Omni and Molmo2 models ([Video & Audio Inputs](/guides/video-audio-inputs)) - * **AWS S3 Bucket Integration:** BYOB dataset storage for training via OIDC federation ([Secure Training](/fine-tuning/finetuning-intro#training-security)) + * **AWS S3 Bucket Integration:** BYOB dataset storage for training via OIDC federation ([Secure Training](/guides/security_compliance/secure_training)) * **Rate Limits Clarification:** Expanded documentation on adaptive serverless limits, upper bounds, and monitoring ([Serverless rate limits](/serverless/rate-limits)) * **Anthropic-Compatible Thinking Parameter:** Control reasoning with `thinking` parameter alongside `reasoning_effort` ([Reasoning](/guides/reasoning)) * **Scaling from Zero Behavior:** Deployments scaled to zero return `503` immediately with retry guidance ([Autoscaling](/deployments/autoscaling)) @@ -134,6 +226,8 @@ + Training Platform + # Warm-Start Training and Azure Model Uploads ## **Warm-Start Training for Reinforcement Fine-Tuning** @@ -171,6 +265,8 @@ + Inference Training Platform + # Playground Categories, New User Roles, Fine-Tuning Improvements, and New Models ## **Playground Categories** @@ -229,6 +325,8 @@ + Inference Training Platform + # Reasoning Guide, Prompt Caching Updates, New Models and CLI Updates ## **Reasoning Guide** @@ -291,6 +389,8 @@ + Inference Training Platform + # DeepSeek V3.2 on Serverless, Cached Token Pricing, and New Models ## ☁️ Serverless @@ -348,6 +448,8 @@ + Inference Training Platform + # Audit Logs, Dataset Download, Weighted Training for Reinforcement Fine-Tuning, and New Model ## **Audit Logs in Web App** @@ -398,6 +500,8 @@ + Inference Training Platform + # Evaluator Improvements, Kimi K2 Thinking on Serverless, and New API Endpoints ## **Improved Evaluator Creation Experience** @@ -463,6 +567,8 @@ + Training + ## ☀️ Sunsetting Build SDK The Build SDK is being deprecated in favor of a new Python SDK generated @@ -490,6 +596,8 @@ + Training + ## Supervised Fine-Tuning We now support supervised fine tuning with separate thinking traces for reasoning models (e.g. DeepSeek R1, GPT OSS, Qwen3 Thinking etc) that ensures training-inference consistency. An example including thinking traces would look like: @@ -516,12 +624,16 @@ + Training + ## Supervised Fine-Tuning We now support Qwen3 MoE model (Qwen3 dense models are already supported) and GPT OSS models for supervised fine-tuning. GPT OSS model fine tunning support is single-turn without thinking traces at the moment. + Training Platform + ## 🎨 Vision-Language Model Fine-Tuning You can now fine-tune Vision-Language Models (VLMs) on Fireworks AI using the Qwen 2.5 VL model family. @@ -540,7 +652,7 @@ * Support for up to 64K context length during training * Built on the same Supervised Fine-tuning V2 infrastructure as text models - See the [VLM fine-tuning documentation](/fine-tuning/fine-tuning-models#vision-fine-tuning) for setup instructions and dataset formatting requirements. + See the [VLM fine-tuning documentation](/fine-tuning/fine-tuning-models#vision-training) for setup instructions and dataset formatting requirements. ## 🔧 Build SDK: Deployment Configuration Application Requirement @@ -566,6 +678,8 @@ + Training + ## 🚀 Bring Your Own Rollout and Reward Development for Reinforcement Learning You can now develop your own custom rollout and reward functionality while using @@ -577,12 +691,16 @@ + Training + ## Supervised Fine-Tuning V2 We now support Llama 4 MoE model supervised fine-tuning (Llama 4 Scout, Llama 4 Maverick, Text only). + Platform + ## 🏗️ Build SDK `LLM` Deployment Logic Refactor Based on early feedback from users and internal testing, we've refactored the @@ -599,6 +717,8 @@ + Inference + ## 🚀 Support for Responses API in Python SDK You can now use the Responses API in the Python SDK. This is useful if you want to use the Responses API in your own applications. @@ -607,6 +727,8 @@ + Platform + ## Support for LinkedIn authentication You can now log in to Fireworks using your LinkedIn account. This is useful if @@ -627,6 +749,8 @@ + Inference Platform + ## Support for GitHub authentication You can now log in to Fireworks using your GitHub account. This is useful if @@ -653,6 +777,8 @@ + Inference Training + ## 🎯 Build SDK: Reward-kit integration for evaluator development The Build SDK now natively integrates with [reward-kit](https://github.com/fw-ai-external/reward-kit) to simplify evaluator development for [Reinforcement Fine-Tuning (RFT)](/fine-tuning/reinforcement-fine-tuning-models). You can now create custom evaluators in Python with automatic dependency management and seamless deployment to Fireworks infrastructure. @@ -677,6 +803,8 @@ + Training + ## Supervised Fine-Tuning V2 Supervised Fine-Tuning V2 released. @@ -696,12 +824,16 @@ + Inference + ## Diarization and batch processing support added to audio inference See our [blog post](https://fireworks.ai/blog/audio-summer-updates-and-new-features) for details. + Training + ## 🚀 Easier & faster LoRA fine-tune deployments on Fireworks You can now deploy a LoRA fine-tune with a single command and get speeds that approximately match the base model: diff --git a/snapshots/google/deprecations.md b/snapshots/google/deprecations.md index decc3ab..63cb110 100644 --- a/snapshots/google/deprecations.md +++ b/snapshots/google/deprecations.md @@ -10,7 +10,8 @@ Deprecation announcements are made on the [Release notes](/gemini-api/docs/chang | **Model** | **Release date** | **Shutdown date** | **Recommended replacement** | | ------------------------------ | ----------------- | -------------------------- | --------------------------- | -| gemini-3.7-flash | August 2026 | No shutdown date announced | | +| gemini-3.8-flash | September 2, 2026 | No shutdown date announced | | +| gemini-3.7-flash | August 13, 2026 | No shutdown date announced | | | gemini-3.6-flash | July 21, 2026 | No shutdown date announced | | | gemini-3.5-flash-lite | July 21, 2026 | No shutdown date announced | | | gemini-3.5-flash | May 19, 2026 | No shutdown date announced | | @@ -65,8 +66,10 @@ Deprecation announcements are made on the [Release notes](/gemini-api/docs/chang | **Model** | **Release date** | **Shutdown date** | **Recommended replacement** | | --------------------------------------------- | ----------------- | -------------------------- | ----------------------------- | +| gemini-3.5-transcribe-live | August 2026 | No shutdown date announced | | | gemini-2.0-flash-live-001 | April 9, 2025 | December 9, 2025 | gemini-3.1-flash-live-preview | | Preview models | | | | +| gemini-3.5-live-translate-preview | June 2026 | No shutdown date announced | | | gemini-3.1-flash-live-preview | March 11, 2026 | No shutdown date announced | | | gemini-2.5-flash-native-audio-preview-12-2025 | December 12, 2025 | No shutdown date announced | gemini-3.1-flash-live-preview | | gemini-live-2.5-flash-preview | June 17, 2025 | December 9, 2025 | gemini-3.1-flash-live-preview | @@ -75,6 +78,7 @@ Deprecation announcements are made on the [Release notes](/gemini-api/docs/chang | **Model** | **Release date** | **Shutdown date** | **Recommended replacement** | | ---------------------------- | ---------------- | -------------------------- | ---------------------------- | +| gemini-3.5-transcribe | August 2026 | No shutdown date announced | | | Preview models | | | | | gemini-3.1-flash-tts-preview | April 13, 2026 | No shutdown date announced | | | gemini-2.5-flash-preview-tts | May 20, 2025 | No shutdown date announced | gemini-3.1-flash-tts-preview | @@ -120,13 +124,22 @@ Deprecation announcements are made on the [Release notes](/gemini-api/docs/chang | veo-3.0-generate-preview | July 31, 2025 | November 12, 2025 | veo-3.1-generate-preview | | veo-3.0-fast-generate-preview | July 31, 2025 | November 12, 2025 | veo-3.1-fast-generate-preview | +## Gemini Omni Flash models + +| **Model** | **Release date** | **Shutdown date** | **Recommended replacement** | +| ------------------------- | ---------------- | -------------------------- | --------------------------- | +| gemini-omni-1.1-flash | August 27, 2026 | No shutdown date announced | | +| Deprecated models | | | | +| gemini-omni-flash-preview | June 30, 2026 | September 30, 2026 | gemini-omni-1.1-flash | + ## Lyria models -| **Model** | **Release date** | **Shutdown date** | **Recommended replacement** | -| -------------------- | ---------------- | -------------------------- | --------------------------- | -| lyria-3-clip-preview | March 25, 2026 | No shutdown date announced | | -| lyria-3-pro-preview | March 25, 2026 | No shutdown date announced | | -| lyria-realtime-exp | May 20, 2025 | No shutdown date announced | | +| **Model** | **Release date** | **Shutdown date** | **Recommended replacement** | +| -------------------- | ----------------- | -------------------------- | --------------------------- | +| lyria-3.5 | September 3, 2026 | No shutdown date announced | | +| lyria-3-clip-preview | March 25, 2026 | No shutdown date announced | | +| lyria-3-pro-preview | March 25, 2026 | No shutdown date announced | lyria-3.5 | +| lyria-realtime-exp | May 20, 2025 | No shutdown date announced | | ## Robotics models diff --git a/snapshots/minimax/api-overview.md b/snapshots/minimax/api-overview.md index bbd2a4a..eadad84 100644 --- a/snapshots/minimax/api-overview.md +++ b/snapshots/minimax/api-overview.md @@ -16,9 +16,9 @@ *** -## LLM +## Large Language Model -The LLM API uses **MiniMax M3**, **MiniMax M2.7**, **MiniMax M2.7 highspeed**, **MiniMax M2.5**, **MiniMax M2.5 highspeed**, **MiniMax M2.1**, **MiniMax M2.1 highspeed**, and **MiniMax M2** to generate conversational content and trigger tool calls based on the provided context. +The Large Language Model API uses **MiniMax M3**, **MiniMax M2.7**, **MiniMax M2.7 highspeed**, **MiniMax M2.5**, **MiniMax M2.5 highspeed**, **MiniMax M2.1**, **MiniMax M2.1 highspeed**, and **MiniMax M2** to generate conversational content and trigger tool calls based on the provided context. It can be accessed via **HTTP requests**, the **Anthropic SDK** (Recommended), or the **OpenAI SDK**. @@ -26,7 +26,7 @@ It can be accessed via **HTTP requests**, the **Anthropic SDK** (Recommended), o | Model Name | Context Window | Description | | :--------------------- | :------------- | :-------------------------------------------------------------------------------------------------------------------------------------------- | -| MiniMax-M3 | 1,000,000 | **Latest M-series language model for agentic reasoning, tool use, coding, and long-context tasks** | +| MiniMax-M3 | 1,000,000 | **Latest M-series language model for agentic reasoning, tool use, coding, and long-context tasks** (output speed approximately 100+ tps) | | MiniMax-M2.7 | 204,800 | **Beginning the journey of recursive self-improvement. (output speed approximately 60 tps)** | | MiniMax-M2.7-highspeed | 204,800 | **M2.7 highspeed: Same performance, faster and more agile (output speed approximately 100 tps)** | | MiniMax-M2.5 | 204,800 | **Peak Performance. Ultimate Value. Master the Complex (output speed approximately 60 tps)** | @@ -49,19 +49,20 @@ Please note: The maximum token count refers to the total number of input and out *** -## MiniMax-H3 \ +## Video Model -This API is powered by MiniMax-H3 and supports video generation from multimodal input (text, images, video, audio), covering text-to-video, image-to-video, first-and-last-frame, and reference-to-video scenarios. +This API supports video generation from multimodal input (text, images, video, audio), covering text-to-video, image-to-video, first-and-last-frame, and reference-to-video scenarios. **Supported Models** -| Model | Description | -| :--------- | :---------------------------------------------------------------------------------------------------------------------------------------- | -| MiniMax-H3 | Multimodal video generation model supporting text / image / first-and-last-frame / reference input, 768P / 2K resolution, 4–15s duration. | +| Model | Description | +| :------------- | :---------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | +| MiniMax-H3 | Multimodal video generation model supporting text / image / first-and-last-frame / reference input, 768P / 2K resolution, 4–15s duration. | +| MiniMax-H3-Max | Fast generation model. Supports text-to-video and image-to-video (first / last frame) only; reference input is not supported. 480P / 768P resolution (no 2K), 5–15s duration. | **API Usage Guide** -MiniMax-H3 tasks are asynchronous. There are three creation endpoints—**Create Video Generation Task**, **Create H3-Context-IR Task**, and **Create Video Regeneration Task**—and shared endpoints for querying, listing, and cancelling or deleting tasks. The workflow is as follows: +Both models share the same `content[]` request protocol and query endpoints — switching models only requires changing the `model` field. MiniMax-H3 tasks are asynchronous. There are three creation endpoints—**Create Video Generation Task**, **Create H3-Context-IR Task**, and **Create Video Regeneration Task**—and shared endpoints for querying, listing, and cancelling or deleting tasks; MiniMax-H3-Max supports the **Create Video Generation Task** endpoint only. The workflow is as follows: 1. Create a video generation task, create an H3-Context-IR task with the same multimodal input, or create a video regeneration task for a source video that meets the MiniMax-H3 768P output specifications. A regeneration request must contain exactly one source-video item with `role=base_video`. All three endpoints return a `task_id` on success. 2. Use **Query Task** with the `task_id` to retrieve its status and result. When a video task succeeds, get its output URL from `content.url`; when an H3-Context-IR task succeeds, get the enhanced prompt from `content.prompt`. You can also use **List Tasks** and distinguish `generation`, `h3_context_ir`, and `regeneration` with `task_type`. @@ -95,21 +96,11 @@ MiniMax-H3 tasks are asynchronous. There are three creation endpoints—**Create *** -## Text to Speech +## Speech Model -This API provides synchronous text-to-speech (T2A) generation, supporting up to **10,000** characters per request. -The interface is stateless: each call only processes the provided input without involving business logic, and the model does not store any user data. +The speech models provide **speech synthesis**, **voice cloning**, and **voice design**, supporting 40 languages and 300+ system voices, with synchronous or asynchronous generation. -**Key Features** - -1. Access to 300+ system voices and custom cloned voices. -2. Adjustable volume, pitch, speed, and output formats. -3. Support for proportional audio mixing. -4. Configurable fixed time intervals. -5. Multiple audio formats and specifications supported: `mp3`, `pcm`, `flac`, `wav`. -6. Support for streaming output. - -**Typical Use Cases:** short text generation, voice chat, online social interactions. +All interfaces are stateless: each call only processes the provided input, does not store user data, and involves no business-logic state. **Supported Models** @@ -122,33 +113,37 @@ The interface is stateless: each call only processes the provided input without | speech-02-hd | Superior rhythm and stability, with outstanding performance in replication similarity and sound quality. | | speech-02-turbo | Superior rhythm and stability, with enhanced multilingual capabilities and excellent performance. | -**Available Interfaces** - -Synchronous speech synthesis provides two interfaces. Choose based on your needs: +**API Overview** -* HTTP T2A API -* WebSocket T2A API +Four capabilities share the models above: -### Supported Languages +1. **Synchronous speech synthesis (T2A)**: real-time text-to-speech, up to **10,000 characters** per request; 300+ system and cloned voices, adjustable volume / pitch / speed, proportional mixing, streaming output, and `mp3` / `pcm` / `flac` / `wav` formats. Available over **HTTP** and **WebSocket**. +2. **Asynchronous long-text synthesis**: up to **1 million characters** per request, ideal for entire books; supports sentence-level timestamps (subtitles). Create a task to get a `task_id`, then use the returned `file_id` with the File API to download (the download URL is valid for **9 hours**). +3. **Voice cloning**: upload the audio to clone to get a `file_id` (optionally upload sample audio to improve quality), then call the cloning API to produce a custom `voice_id`. Individual or enterprise verification is required. +4. **Voice design**: generate a personalized voice from a description prompt; the resulting `voice_id` can be used directly with the synthesis APIs above. -MiniMax speech synthesis models offer robust multilingual capability, supporting **40 widely used languages** worldwide. + + Voices produced by cloning and voice design are **temporary**: the fee is charged only on first use in speech synthesis (previews within those APIs do not count). If the voice is not used by any speech synthesis API within **168 hours (7 days)**, it is deleted. + -| Support Languages | | | -| ----------------- | ------------- | ------------- | -| 1. Chinese | 15. Turkish | 28. Malay | -| 2. Cantonese | 16. Dutch | 29. Persian | -| 3. English | 17. Ukrainian | 30. Slovak | -| 4. Spanish | 18. Thai | 31. Swedish | -| 5. French | 19. Polish | 32. Croatian | -| 6. Russian | 20. Romanian | 33. Filipino | -| 7. German | 21. Greek | 34. Hungarian | -| 8. Portuguese | 22. Czech | 35. Norwegian | -| 9. Arabic | 23. Finnish | 36. Slovenian | -| 10. Italian | 24. Hindi | 37. Catalan | -| 11. Japanese | 25. Bulgarian | 38. Nynorsk | -| 12. Korean | 26. Danish | 39. Tamil | -| 13. Indonesian | 27. Hebrew | 40. Afrikaans | -| 14. Vietnamese | | | + + | Support Languages | | | + | ----------------- | ------------- | ------------- | + | 1. Chinese | 15. Turkish | 28. Malay | + | 2. Cantonese | 16. Dutch | 29. Persian | + | 3. English | 17. Ukrainian | 30. Slovak | + | 4. Spanish | 18. Thai | 31. Swedish | + | 5. French | 19. Polish | 32. Croatian | + | 6. Russian | 20. Romanian | 33. Filipino | + | 7. German | 21. Greek | 34. Hungarian | + | 8. Portuguese | 22. Czech | 35. Norwegian | + | 9. Arabic | 23. Finnish | 36. Slovenian | + | 10. Italian | 24. Hindi | 37. Catalan | + | 11. Japanese | 25. Bulgarian | 38. Nynorsk | + | 12. Korean | 26. Danish | 39. Tamil | + | 13. Indonesian | 27. Hebrew | 40. Afrikaans | + | 14. Vietnamese | | | + @@ -158,86 +153,15 @@ MiniMax speech synthesis models offer robust multilingual capability, supporting Streaming speech synthesis via WebSocket - - -*** - -## Asynchronous Long-Text Speech Generation - -This API supports asynchronous text-to-speech generation. Each request can handle up to **1 million characters**, and the resulting audio can be retrieved asynchronously. -Features supported: - -1. Choose from 100+ system voices and cloned voices. -2. Customize pitch, speed, volume, bitrate, sample rate, and output format. -3. Retrieve audio metadata, such as duration and file size. -4. Retrieve precise sentence-level timestamps (subtitles). -5. Input text directly as a string or via `file_id` after uploading a text file. -6. Detect illegal characters: - * If illegal characters are **≤10%**, audio is generated normally, with the ratio returned. - * If illegal characters are **>10%**, no audio will be generated (an error code will be returned). - -**Note:** The returned audio URL is valid for **9 hours** (32,400 seconds) from the time it is issued. After expiration, the URL becomes invalid and the generated data will be lost. - -**Use Case:** Converting entire books or other long texts into audio. - -**Supported Models** - -| Model | Description | -| :--------------- | :------------------------------------------------------------------------------------------------------- | -| speech-2.8-hd | Latest HD model. Ultra-realistic quality featuring sound tags. | -| speech-2.8-turbo | Latest Turbo model. Seamless speed meets natural flow. | -| speech-2.6-hd | HD model with outstanding prosody and excellent cloning similarity. | -| speech-2.6-turbo | Turbo model with support for 40 languages. | -| speech-02-hd | Superior rhythm and stability, with outstanding performance in replication similarity and sound quality. | -| speech-02-turbo | Superior rhythm and stability, with enhanced multilingual capabilities and excellent performance. | - -**API Overview** - -This feature includes **two APIs**: - -1. Create a speech generation task (returns `task_id`). -2. Query the speech generation task status using `task_id`. -3. If the task succeeds, use the returned `file_id` with the **File API** to view and download the result. - - Create a long-text speech generation task - + Query speech generation task status - - -*** - -## Voice Cloning - -This API supports cloning voices from user-uploaded audio files along with optional sample audio to enhance cloning quality. - -**Use cases:** fast replication of a target timbre (IP voice recreation, voice cloning) where you need to quickly clone a specific voice. - -The API supports cloning from mono or stereo audio and can rapidly reproduce speech that matches the timbre of a provided reference file. -**Supported Models** - -| Model | Description | -| :--------------- | :------------------------------------------------------------------------------------------------------- | -| speech-2.8-hd | Latest HD model. Ultra-realistic quality featuring sound tags. | -| speech-2.8-turbo | Latest Turbo model. Seamless speed meets natural flow. | -| speech-2.6-hd | HD model with real-time response, intelligent parsing, fluent LoRA voice | -| speech-2.6-turbo | Turbo model. Ultimate Value, 40 Languages | -| speech-02-hd | Superior rhythm and stability, with outstanding performance in replication similarity and sound quality. | -| speech-02-turbo | Superior rhythm and stability, with enhanced multilingual capabilities and excellent performance. | - -### Notes - -* Using this API to clone a voice **does not** immediately incur a cloning fee. The cloning fee is charged the **first time** you synthesize speech with the cloned voice in a T2A synthesis API (the preview/audition within this API does not count). -* Voices produced via this rapid cloning API are **temporary**. To keep a cloned voice permanently, call **any** T2A speech synthesis API with that voice **within 168 hours (7 days)** (the preview/audition within this API does not count). If the time limit is exceeded, the voice will be deleted. -* This API is stateless: each call only processes the incoming data, does not store user-uploaded content, and involves no business-logic state. - - Upload audio file to clone @@ -245,37 +169,11 @@ The API supports cloning from mono or stereo audio and can rapidly reproduce spe Execute voice cloning - - -*** - -## Voice Design - -This API supports generating personalized custom voices based on user-provided voice description prompts. - -The generated voices (voice\_id) can then be used in the T2A API and the T2A Async API for speech generation. -**Supported Models** - -> It is recommended to use **speech-02-hd** for the best results. - -| Model | Description | -| :--------------- | :------------------------------------------------------------------------------------------------------- | -| speech-2.8-hd | Latest HD model. Ultra-realistic quality featuring sound tags. | -| speech-2.8-turbo | Latest Turbo model. Seamless speed meets natural flow. | -| speech-2.6-hd | HD model with real-time response, intelligent parsing, fluent LoRA voice | -| speech-2.6-turbo | Turbo model. Ultimate Value, 40 Languages | -| speech-02-hd | Superior rhythm and stability, with outstanding performance in replication similarity and sound quality. | -| speech-02-turbo | Superior rhythm and stability, with enhanced multilingual capabilities and excellent performance. | - -### Notes - -> * Using this API to generate a voice does not immediately incur a fee. The generation fee will be charged upon the first use of the generated voice in speech synthesis. -> * Voices generated through this API are temporary. If you wish to keep a voice permanently, you must use it in any speech synthesis API within 168 hours (7 days). - - - Generate personalized voices from descriptions - + + Generate personalized voices from descriptions + + *** @@ -349,7 +247,13 @@ Supported file formats, capacity, and size limits are defined by the **Upload Fi *** -## Official MCP +## Tools + +**Web Search** + +`web_search` is a server-side web search tool hosted and executed by MiniMax. The model can retrieve up-to-date information while generating a response and answer based on the search results. It is available through both the Anthropic Messages API and the OpenAI Responses API. See [Web Search](/docs/guides/server-tools#web_search) for interface details and examples. + +**Official MCP** MiniMax provides official Model Context Protocol (MCP) server implementations: diff --git a/snapshots/moonshot/models.md b/snapshots/moonshot/models.md index 1b49f32..d51b326 100644 --- a/snapshots/moonshot/models.md +++ b/snapshots/moonshot/models.md @@ -4,13 +4,13 @@ # Model List -> Review currently available Kimi multimodal, coding, and Moonshot V1 models, plus migration guidance for discontinued models. +> Review currently available Kimi multimodal and coding models, plus migration guidance for discontinued models. Click [here](pricing/chat) to see more details of model price. - - Following the Kimi K3 launch, `kimi-k2.5` and the `moonshot-v1` series are no longer available to newly registered users (full platform sunset on August 31). Please switch to a newer model as soon as possible. - + + The `kimi-k2.5` and `moonshot-v1` series were officially retired on August 31, 2026. Calls to these models now return a 404 (model not found) error. Please migrate to [kimi-k3](/docs/guide/kimi-k3-quickstart) or other latest models. + ## Multi-modal Model @@ -20,32 +20,30 @@ Click [here](pricing/chat) to see more details of model price. | `kimi-k2.7-code` | Kimi's dedicated coding model. It follows instructions more reliably in long contexts, completes coding tasks with higher success rates. Context 256k | | `kimi-k2.7-code-highspeed` | High-Speed version of Kimi K2.7 Code model, with output speed of approximately 180 Tokens/s and up to 260 Tokens/s in short context scenarios, delivering a more extreme coding experience. | | `kimi-k2.6` | Supports both visual and text input, thinking and non-thinking modes, and dialogue and Agent tasks. Context 256k | -| `kimi-k2.5` | Achieves open-source SoTA performance in Agent, code, visual understanding, and a range of general intelligent tasks. It also supports visual and text input, thinking and non-thinking modes, and dialogue and Agent tasks. Context 256k | -## Generation Model Moonshot V1 - -| Model Name | Description | -| --------------------------------- | ----------------------------------------------------------------------------- | -| `moonshot-v1-8k` | Suitable for generating short texts, context length 8k | -| `moonshot-v1-32k` | Suitable for generating long texts, context length 32k | -| `moonshot-v1-128k` | Suitable for generating very long texts, context length 128k | -| `moonshot-v1-8k-vision-preview` | Vision model, understands image content and outputs text, context length 8k | -| `moonshot-v1-32k-vision-preview` | Vision model, understands image content and outputs text, context length 32k | -| `moonshot-v1-128k-vision-preview` | Vision model, understands image content and outputs text, context length 128k | +## Deprecated Models -> Note: The only difference between these Moonshot V1 models is their maximum context length (including input and output), there is no difference in effect. +> `kimi-k2.5` was officially discontinued on **August 31, 2026** and is no longer maintained or supported. Please use the latest Kimi model [kimi-k3](/docs/guide/kimi-k3-quickstart) for continued support and enhanced reasoning capabilities. -## Deprecated Models +> The `moonshot-v1` series models (including `moonshot-v1-auto` and the `-vision-preview` variants) were officially discontinued on **August 31, 2026** and are no longer maintained or supported. Please use the latest Kimi model [kimi-k3](/docs/guide/kimi-k3-quickstart) for continued support and enhanced reasoning capabilities. > The `kimi-k2` series models were officially discontinued on **May 25, 2026** and are no longer maintained or supported. Please use the latest Kimi model [kimi-k3](/docs/guide/kimi-k3-quickstart) for continued support and enhanced reasoning capabilities. -| Model Name | Description | -| ------------------------ | ----------- | -| `kimi-k2-0905-preview` | Deprecated | -| `kimi-k2-0711-preview` | Deprecated | -| `kimi-k2-turbo-preview` | Deprecated | -| `kimi-k2-thinking` | Deprecated | -| `kimi-k2-thinking-turbo` | Deprecated | +| Model Name | Description | +| --------------------------------- | ----------- | +| `kimi-k2.5` | Deprecated | +| `moonshot-v1-8k` | Deprecated | +| `moonshot-v1-32k` | Deprecated | +| `moonshot-v1-128k` | Deprecated | +| `moonshot-v1-auto` | Deprecated | +| `moonshot-v1-8k-vision-preview` | Deprecated | +| `moonshot-v1-32k-vision-preview` | Deprecated | +| `moonshot-v1-128k-vision-preview` | Deprecated | +| `kimi-k2-0905-preview` | Deprecated | +| `kimi-k2-0711-preview` | Deprecated | +| `kimi-k2-turbo-preview` | Deprecated | +| `kimi-k2-thinking` | Deprecated | +| `kimi-k2-thinking-turbo` | Deprecated | > `kimi-latest` was officially discontinued on **January 28, 2026** and is no longer maintained or supported. Please use the latest Kimi model [kimi-k3](/docs/guide/kimi-k3-quickstart) for continued support and enhanced reasoning capabilities. diff --git a/snapshots/openai/deprecations.md b/snapshots/openai/deprecations.md index 7efef0a..70df17d 100644 --- a/snapshots/openai/deprecations.md +++ b/snapshots/openai/deprecations.md @@ -34,6 +34,19 @@ We use the term "legacy" to refer to models and endpoints that no longer receive Upcoming deprecations are listed below, with the most recent announcements at the top. +### 2026-08-26: Transcription models + +On August 26, 2026, we notified developers using `whisper-1`, `gpt-4o-transcribe`, `gpt-4o-mini-transcribe`, and `gpt-4o-transcribe-diarize` of their deprecation and removal from the API on February 26, 2027. + +For information about the recommended replacements, see the [transcription guide](https://developers.openai.com/api/docs/guides/transcription). + +| Shutdown date | Model / system | Recommended replacement | +| ------------- | --------------------------- | ----------------------------------------- | +| Feb 26, 2027 | `whisper-1` | `gpt-live-transcribe` or `gpt-transcribe` | +| Feb 26, 2027 | `gpt-4o-transcribe` | `gpt-live-transcribe` or `gpt-transcribe` | +| Feb 26, 2027 | `gpt-4o-mini-transcribe` | `gpt-live-transcribe` or `gpt-transcribe` | +| Feb 26, 2027 | `gpt-4o-transcribe-diarize` | `gpt-live-transcribe` or `gpt-transcribe` | + ### 2026-07-20: Legacy audio, realtime, and transcription models On July 20, 2026, we notified developers using legacy audio, realtime, and transcription model families and snapshots of their deprecation and removal from the API on January 20, 2027. @@ -174,18 +187,6 @@ To improve reliability and make it easier for developers to choose the right mod | 2026-09-28 | `davinci-002` | `gpt-5.6-terra` | | 2026-09-28 | `gpt-3.5-turbo-1106` | `gpt-5.6-terra` | -### 2025-08-20: Assistants API - -On August 26th, 2025, we notified developers using the Assistants API of its deprecation and removal from the API one year later, on August 26, 2026. - -When we released the [Responses API](https://developers.openai.com/api/reference/resources/responses/methods/create) in [March 2025](https://developers.openai.com/api/docs/changelog), we announced plans to bring all Assistants API features to the easier to use Responses API, with a sunset date in 2026. - -See the Assistants to Conversations [migration guide](https://developers.openai.com/api/docs/assistants/migration) to learn more about how to migrate your current integration to the Responses API and Conversations API. - -| Shutdown date | Model / system | Recommended replacement | -| ------------- | -------------- | ----------------------------------- | -| 2026‑08‑26 | Assistants API | Responses API and Conversations API | - ## Past deprecations Past deprecations are listed below, with the most recent announcements at the top. @@ -280,6 +281,18 @@ In September, 2025, we notified developers using gpt-4o-realtime-preview models | 2026-05-07 | gpt-4o-audio-preview | gpt-audio-1.5 | | 2026-05-07 | gpt-4o-mini-audio-preview | gpt-audio-mini | +### 2025-08-20: Assistants API + +On August 26th, 2025, we notified developers using the Assistants API of its deprecation and removal from the API one year later, on August 26, 2026. + +When we released the [Responses API](https://developers.openai.com/api/reference/resources/responses/methods/create) in [March 2025](https://developers.openai.com/api/docs/changelog), we announced plans to bring all Assistants API features to the easier to use Responses API, with a sunset date in 2026. + +See the Assistants to Conversations [migration guide](https://developers.openai.com/api/docs/assistants/migration) to learn more about how to migrate your current integration to the Responses API and Conversations API. + +| Shutdown date | Model / system | Recommended replacement | +| ------------- | -------------- | ----------------------------------- | +| 2026‑08‑26 | Assistants API | Responses API and Conversations API | + ### 2025-06-10: gpt-4o-realtime-preview-2024-10-01 On June 10th, 2025, we notified developers using gpt-4o-realtime-preview-2024-10-01 of its deprecation and removal from the API in three months. diff --git a/snapshots/perplexity/changelog.md b/snapshots/perplexity/changelog.md index 3b5f008..c921e0f 100644 --- a/snapshots/perplexity/changelog.md +++ b/snapshots/perplexity/changelog.md @@ -8,6 +8,18 @@ Looking ahead? Check out our [Feature Roadmap](/docs/resources/feature-roadmap) to see what's coming next. + + **GLM 5.3 Flash** + + The Agent API and Router API now support `perplexity/glm-5.3-flash` at \$0.15 per million uncached-input tokens, \$0.03 per million cached-input tokens, and \$0.50 per million output tokens. See the [Agent API Models reference](/docs/agent-api/models) or the [Router model catalog](/docs/router/models). + + + + **GLM 5.3** + + The Agent API and Router API now support `perplexity/glm-5.3` at \$1.40 per million uncached-input tokens, \$0.26 per million cached-input tokens, and \$4.40 per million output tokens. See the [Agent API Models reference](/docs/agent-api/models) or the [Router model catalog](/docs/router/models). + + **Prompt caching for presets** @@ -23,7 +35,7 @@ **Gemini 3.7 Flash** - The Agent API and Router API now support `google/gemini-3.7-flash` at launch pricing of \$0.375 per million input tokens, \$0.0375 per million cached-input tokens, and \$1.875 per million output tokens. See the [Agent API Models reference](/docs/agent-api/models). + Pricing for `google/gemini-3.7-flash` increased on August 27, 2026 to \$0.75 per million input tokens, \$0.075 per million cached-input tokens, and \$3.75 per million output and reasoning tokens. See the [Agent API Models reference](/docs/agent-api/models). @@ -35,19 +47,19 @@ **NVIDIA Nemotron 3 Ultra** - The Agent API and Router API now support `perplexity/nemotron-3-ultra-550b-a55b` at \$0.25 per million input or cached-input tokens and \$2.50 per million output tokens. See the [Agent API Models reference](/docs/agent-api/models) or the [Router model catalog](/docs/gateway/models). + The Agent API and Router API now support `perplexity/nemotron-3-ultra-550b-a55b` at \$0.25 per million input or cached-input tokens and \$2.50 per million output tokens. See the [Agent API Models reference](/docs/agent-api/models) or the [Router model catalog](/docs/router/models). **NVIDIA Nemotron 3.5 Lightning** - The Agent API and Router API now support `perplexity/nemotron-3.5-lightning-30b-a3b`, a fast, efficient open-weight reasoning model, at \$0.0115 per million input tokens, \$0.00115 per million cached-input tokens, and \$0.17 per million output tokens. See the [Agent API Models reference](/docs/agent-api/models) or the [Router model catalog](/docs/gateway/models). + The Agent API and Router API now support `perplexity/nemotron-3.5-lightning-30b-a3b`, a fast, efficient open-weight reasoning model, at \$0.0115 per million input tokens, \$0.00115 per million cached-input tokens, and \$0.17 per million output tokens. See the [Agent API Models reference](/docs/agent-api/models) or the [Router model catalog](/docs/router/models). **DeepSeek V4 Flash 0731** - The Agent API and Router API now support `perplexity/deepseek-v4-flash-0731`, a fast, efficient open reasoning model with a 1M-token context window. See pricing in the [Agent API Models reference](/docs/agent-api/models) or the [Router model catalog](/docs/gateway/models). + The Agent API and Router API now support `perplexity/deepseek-v4-flash-0731`, a fast, efficient open reasoning model with a 1M-token context window. See pricing in the [Agent API Models reference](/docs/agent-api/models) or the [Router model catalog](/docs/router/models). @@ -67,7 +79,7 @@ * Automatic health-based routing and failover across model deployments * Per-token pricing at each model's published rates, with no per-request fees - [Get started with the Router API →](/docs/gateway/quickstart) + [Get started with the Router API →](/docs/router/quickstart) diff --git a/snapshots/thinking-machines/model-deprecations.md b/snapshots/thinking-machines/model-deprecations.md index 6c17d9b..dc8bdd3 100644 --- a/snapshots/thinking-machines/model-deprecations.md +++ b/snapshots/thinking-machines/model-deprecations.md @@ -1,3 +1,5 @@ +[ View as Markdown](index.md "Open the Markdown source of this page") + # Model Deprecations ## How does Tinker handle model deprecation? diff --git a/snapshots/vertex/model-versions.md b/snapshots/vertex/model-versions.md index 04c8475..e8e0a60 100644 --- a/snapshots/vertex/model-versions.md +++ b/snapshots/vertex/model-versions.md @@ -24,17 +24,17 @@ The following table lists the models that will be available for at least 12 mont ### Gemini image models -| Model ID | Release date | Retirement date | Replacement model | -| ---------------------- | --------------- | --------------------- | --------------------------- | -| gemini-3-pro-image | May 28, 2026 | May 28, 2027 or later | | -| gemini-3.1-flash-image | May 28, 2026 | May 28, 2027 or later | | -| gemini-2.5-flash-image | October 2, 2025 | October 2, 2026 | Gemini 3.1 Flash-Lite Image | +| Model ID | Release date | Retirement date | Replacement model | +| --------------------------- | --------------- | ---------------------------- | --------------------------- | +| gemini-3.1-flash-lite-image | June 23, 2026 | No retirement date announced | | +| gemini-3-pro-image | May 28, 2026 | May 28, 2027 or later | | +| gemini-3.1-flash-image | May 28, 2026 | May 28, 2027 or later | | +| gemini-2.5-flash-image | October 2, 2025 | October 2, 2026 | Gemini 3.1 Flash-Lite Image | ### Veo models | Model ID | Release date | Retirement date | Replacement model | | ------------------------- | ----------------- | -------------------------- | ------------------------- | -| veo-2.0-generate-001 | May 27, 2025 | June 30, 2026 | veo-3.1-generate-001 | | veo-3.0-generate-001 | July 29, 2025 | June 30, 2026 | veo-3.1-generate-001 | | veo-3.0-fast-generate-001 | July 29, 2025 | June 30, 2026 | veo-3.1-fast-generate-001 | | veo-3.1-generate-001 | November 17, 2025 | November 17, 2026 or later | | @@ -55,11 +55,11 @@ The following table lists the models that will be available for at least 12 mont Short-term availability models retire 45 days after a replacement model is released. The following table lists models available for shorter terms: -| Model ID | Release date | Retirement date | Replacement model | -| --------------------------- | --------------- | ---------------------------- | ----------------- | -| gemini-3.7-flash | August 13, 2026 | No retirement date announced | | -| gemini-3.6-flash | July 21, 2026 | No retirement date announced | | -| gemini-3.1-flash-lite-image | June 30, 2026 | No retirement date announced | | +| Model ID | Release date | Retirement date | Replacement model | +| ---------------- | ----------------- | ---------------------------- | ----------------- | +| gemini-3.8-flash | September 2, 2026 | No retirement date announced | | +| gemini-3.7-flash | August 13, 2026 | No retirement date announced | | +| gemini-3.6-flash | July 21, 2026 | No retirement date announced | | ### Retired models diff --git a/snapshots/xai/may-15-retirement.md b/snapshots/xai/may-15-retirement.md index 2de8428..4eed39e 100644 --- a/snapshots/xai/may-15-retirement.md +++ b/snapshots/xai/may-15-retirement.md @@ -23,7 +23,7 @@ Starting **May 15, 2026 at 12:00 PM PT**: * Requests to any **reasoning** model in the list above will be served by **`grok-4.3` with `low` reasoning effort**. * Requests to any **non-reasoning** model in the list above will be served by **`grok-4.3` with `none` reasoning effort**. -* `grok-imagine-image-pro` will be redirected to `grok-imagine-image-quality`. +* `grok-imagine-image-pro` will be redirected to `grok-imagine-image-quality`. On November 2, 2026, `grok-imagine-image-quality` is itself retired in favor of `grok-imagine-image-2.0`; see the [November 2 migration guide](/developers/migration/imagine-image-quality-nov-2). ## Pricing impact diff --git a/snapshots/z-ai/pricing.md b/snapshots/z-ai/pricing.md index 63f3bc9..1343d82 100644 --- a/snapshots/z-ai/pricing.md +++ b/snapshots/z-ai/pricing.md @@ -8,17 +8,28 @@ ## Models +### Latest Models + +Prices per 1M tokens. + +| Model | Input | Cached Input | Cached Input Storage | Output | +| :------------ | :----------------- | :----------------- | :------------------- | :---------------- | +| GLM-5.3-Flash | ~~\$0.15~~ \$0.075 | ~~\$0.03~~ \$0.015 | Limited-time Free | ~~\$0.50~~ \$0.25 | +| GLM-5.3 | \$1.4 | \$0.26 | Limited-time Free | \$4.4 | +| GLM-5.2 | \$1.4 | \$0.26 | Limited-time Free | \$4.4 | + + + GLM-5.3-Flash is available at a 50% discount (strikethrough prices are list prices). The promotion ends at 24:00 on September 9, 2026 (UTC+8, Singapore time). + + ### Text Models Prices per 1M tokens. | Model | Input | Cached Input | Cached Input Storage | Output | | :------------------ | :----- | :----------- | :------------------- | :----- | -| GLM-5.3 | \$1.4 | \$0.26 | Limited-time Free | \$4.4 | -| GLM-5.2 | \$1.4 | \$0.26 | Limited-time Free | \$4.4 | | GLM-5.1 | \$1.4 | \$0.26 | Limited-time Free | \$4.4 | | GLM-5 | \$1 | \$0.2 | Limited-time Free | \$3.2 | -| GLM-5-Turbo | \$1.2 | \$0.24 | Limited-time Free | \$4.0 | | GLM-4.7 | \$0.6 | \$0.11 | Limited-time Free | \$2.2 | | GLM-4.7-FlashX | \$0.07 | \$0.01 | Limited-time Free | \$0.4 | | GLM-4.6 | \$0.6 | \$0.11 | Limited-time Free | \$2.2 | @@ -36,7 +47,6 @@ Prices per 1M tokens. | Model | Input | Cached Input | Cached Input Storage | Output | | :-------------- | :----- | :----------- | :------------------- | :----- | -| GLM-5V-Turbo | \$1.2 | \$0.24 | Limited-time Free | \$4 | | GLM-4.6V | \$0.3 | \$0.05 | Limited-time Free | \$0.9 | | GLM-OCR | \$0.03 | \\ | \\ | \$0.03 | | GLM-4.6V-FlashX | \$0.04 | \$0.004 | Limited-time Free | \$0.4 | @@ -62,15 +72,9 @@ Prices per image. Prices per video. -| Model | Price | -| :--------------- | :---- | -| CogVideoX-3 | \$0.2 | -| ViduQ1-Text | \$0.4 | -| ViduQ1-Image | \$0.4 | -| ViduQ1-Start-End | \$0.4 | -| Vidu2-Image | \$0.2 | -| Vidu2-Start-End | \$0.2 | -| Vidu2-Reference | \$0.4 | +| Model | Price | +| :---------- | :---- | +| CogVideoX-3 | \$0.2 | ### Audio Models diff --git a/snapshots/z-ai/releases.md b/snapshots/z-ai/releases.md index 4a167f0..ab92c14 100644 --- a/snapshots/z-ai/releases.md +++ b/snapshots/z-ai/releases.md @@ -8,6 +8,14 @@ ## Models + + * Native visual capabilities enable the model to observe interfaces, rendering results, and interaction feedback—creating a closed loop across code, browsers, and GUIs. + + * Efficient hybrid architecture: Combines linear and sparse attention with 320B total parameters and 18B activated, significantly reducing compute and KV-cache requirements. + + * Beyond coding: Supports office document and financial research workflows, autonomously breaking down goals, using tools, and refining outputs. Learn more in our [documentation](/guides/vlm/glm-5.3-flash).\* + + * Stronger Coding Capabilities: GLM-5.3 delivers a significant improvement in coding capabilities, achieving a 50% gain over GLM-5.2 on Z.ai Code Bench and reaching state-of-the-art (SOTA) performance among open-source models on public benchmarks, including Terminal Bench 3.0. @@ -28,18 +36,6 @@ * It demonstrates stronger engineering intelligence across autonomous planning, sustained execution, bug fixing, and strategy iteration, while achieving comprehensive capability alignment with Claude Opus 4.6. Built with multi-turn SFT, RL, and a process-quality evaluation framework, GLM-5.1 further improves stability, consistency, and tool use over extended tasks. Learn more in our [documentation](/guides/llm/glm-5.1).\* - - * Designed for vision-based coding and claw-style Agent workflows, GLM-5V-Turbo brings native multimodal understanding to images, video, and text, enabling stronger planning, coding, and action execution in visually grounded tasks. - - * It strengthens multimodal tool use and GUI task execution, allowing the model to better interpret interfaces, design mocks, documents, and charts, plan the next action, and complete end-to-end Agent workflows while maintaining strong pure-text coding and reasoning performance. Learn more in our [documentation](/guides/vlm/glm-5v-turbo).\* - - - - * Designed for high-throughput OpenClaw lobster workloads, GLM-5-Turbo focuses on improving stability and efficiency in long-chain Agent tasks, enabling smoother execution for complex, multi-step workflows. - - * It strengthens tool and Skills integration and enhances complex instruction decomposition, allowing the model to better identify task goals, plan execution steps, coordinate across multiple agents, and maintain temporal consistency in extended tasks.Learn more in our [documentation](/guides/llm/glm-5-turbo).\* - - * Designed for complex system engineering and long-range Agent tasks, GLM-5 shifts the paradigm from coding to engineering, demonstrating strong deep-reasoning performance in backend architecture, complex algorithms, and stubborn bug fixing.