MC

This commit introduces **new capabilities** for **Workers AI models** by explicitly surfacing "reasoning" and "vision" properties within the model catalog and individual model pages. It involves a significant **refactoring** of the **model display components** (`ModelCatalog`, `ModelBadges`, `ModelFeatures`) to utilize a new, centralized `model-properties` utility for a more data-driven approach to capability definitions. This update also adds "Reasoning" and "Vision" rows to the model features information table and updates several model JSONs to include these new properties. Additionally, **NVIDIA** and **Zhipu AI** are integrated as new author logos and catalog entries, enhancing the visual representation of model providers. The changes improve the discoverability of advanced model features and streamline the maintenance of model metadata.

This commit performs a **maintenance update** to the **Workers AI model schemas**, synchronizing them with the latest API specifications. It introduces **new capabilities** for `glm-4.7-flash` by adding `chat_template_kwargs` for thinking toggles, and significantly **enhances** `nemotron-3-120b-a12b` by expanding its context window and adopting a full OpenAI-compatible schema. Additionally, it **refines** several Llama, Mistral, and QWQ models by removing the `tool_call_id` pattern regex and adjusts the `context_window` for `llama-3.2-3b-instruct`, ensuring all listed models reflect their most current and accurate configurations. This update impacts the **`src/content/workers-ai-models`** directory, affecting how these models are described and potentially interacted with by downstream applications.

This commit **introduces a new capability** by adding the **NVIDIA Nemotron 3 Super model** to the **Workers AI platform**. It includes the necessary model definition in `src/content/workers-ai-models/nemotron-3-120b-a12b.json`, enabling users to leverage this powerful new large language model. Concurrently, this update provides comprehensive **documentation** for the new model, including entries in the Workers AI changelog and release notes. Furthermore, the pricing information in `src/content/docs/workers-ai/platform/pricing.mdx` has been updated to reflect the costs associated with the Nemotron 3 Super model, ensuring users are fully informed about its availability and usage.

This commit **updates the code ownership configuration** for the **AI Gateway** and **Workers AI** services. It **adds @zeke, @superhighfives, and @bfirsh** to the `.github/CODEOWNERS` file, assigning them responsibility for pull requests impacting these critical AI-related components. This **maintenance** change ensures proper review and accountability for future modifications to these systems, streamlining development workflows and enhancing code quality.

This commit introduces significant enhancements and fixes to the **Workers AI platform**, primarily by adding **Chat Completions API support** for the `gpt-oss-120b` and `gpt-oss-20b` models, which now handle messages, functions, and tools. It also includes crucial **bug fixes** for 45 other text generation models, updating their input schemas to correctly support both string and array content for messages, thereby improving API flexibility and reliability. Furthermore, the `whisper-large-v3-turbo` model gains new input parameters and `async_queue` support, expanding its capabilities. These **new capabilities** and **schema fixes** are reflected in updated API format documentation and release notes, ensuring users have accurate information for leveraging these improved AI models. This comprehensive update broadens the functionality and strengthens the stability of the **Workers AI** ecosystem.

This commit **introduces the `GLM-4.7-Flash` model** to the **Workers AI platform**, providing its core definition and comprehensive user documentation. It adds the model's JSON definition, featuring an OpenAI-compatible chat-completions schema, and **updates the changelog, release notes, and pricing documentation** to reflect its availability. This work represents a **new capability** for Workers AI users, ensuring they have all necessary information for integrating and understanding the new model. Additionally, minor **maintenance fixes** for broken Workers AI binding links are included.

This commit performs a **maintenance fix** by **correcting a data error** within the project's **release notes** content. Specifically, it updates the `src/content/release-notes/workers-ai.yaml` file, which details information for the **Workers AI** platform. The change rectifies an incorrect year in the changelog link for the `Black Forest Labs FLUX.2` model, updating it from 2025 to 2026. This ensures that users accessing the release notes are directed to the correct and up-to-date information, preventing potential broken links or confusion.

This commit performs a **documentation update** by **correcting a date** within the **Workers AI changelog and release notes**. Specifically, it changes the `publish_date` and file name for the `flux-2-klein-4b` entry from 2025-01-15 to **2026-01-15** in `src/content/changelog/workers-ai/2026-01-15-flux-2-klein-4b-workers-ai.mdx` and `src/content/release-notes/workers-ai.yaml`. This ensures the **public-facing documentation** accurately reflects the intended release timeline for this specific Workers AI model.

This commit introduces the **FLUX.2 [klein] 4B model** as a **new capability** on the **Workers AI platform**, enabling its use and display. It adds the core configuration for the model in `src/content/workers-ai-models/flux-2-klein-4b.json` and updates the model display page `src/pages/workers-ai/models/[name].astro` to feature it. Concurrently, extensive **documentation updates** are provided, including a new changelog entry, release notes, and crucial **pricing information** in `src/content/docs/workers-ai/platform/pricing.mdx`. A minor **refactoring** of the `Flux-2` component also simplifies its internal multipart form data preparation logic.

This commit **introduces pricing for the Deepgram Flux model** within the **Workers AI platform** and performs **refactoring of several existing model definitions**. It adds the necessary pricing configuration for `flux.json`, updates the `pricing.mdx` documentation with details, and includes a release note about the promotional period ending. Concurrently, it **refactors** various `workers-ai-models/*.json` files by simplifying input/output/prompt/message titles and removing an `async queue` property for improved consistency and clarity. This work enhances the **Workers AI model catalog** by standardizing definitions and clearly communicating pricing for a key speech-to-text model.

This commit **introduces new code examples** for the **`Flux-2-Dev` model**, significantly enhancing its documentation. A dedicated **Astro component** (`Flux-2-Dev.astro`) was created to display these examples, which are then **integrated** into the **`workers-ai` model pages** via `[name].astro`. This work represents a **new capability** for the `Flux-2-Dev` model and includes **maintenance** by removing an outdated pricing notice from the `workers-ai` model pages, ensuring up-to-date information for users.

This commit **updates documentation** for the **Workers AI binding**, specifically within a changelog entry, to address a known technical limitation. It introduces a **temporary workaround** to the JavaScript code examples, demonstrating how to correctly handle **multipart requests** when interacting with Workers AI. This **documentation update** provides crucial guidance for developers, ensuring they can successfully implement features requiring multipart data until a more permanent solution is available.

This commit **updates the Workers AI changelog documentation** by adding a JavaScript helper function `streamToBlob` as a practical example. This **documentation enhancement** provides developers with a concrete code snippet, improving the clarity and utility of the changelog entry. Specifically, it modifies the `src/content/changelog/workers-ai/2025-11-25-flux-2-dev-workers-ai.mdx` file to offer a clearer understanding of certain functionalities within the **Workers AI** ecosystem.

This commit **introduces a new capability** by integrating the **Black Forest Labs FLUX.2 [dev] model into Workers AI**. This involves creating a new JSON definition file for the `@cf/black-forest-labs/flux-2-dev` model, outlining its description, task, and schema. Comprehensive **documentation updates** accompany this launch, including a new changelog entry, updated pricing details in `src/content/docs/workers-ai/platform/pricing.mdx`, and a release note entry to inform users of the model's availability.

This commit **synchronizes** the local **Workers AI model definitions** with the upstream API, ensuring the documentation and internal representations are current. It involves extensive **refactoring** of existing model metadata, such as standardizing pricing units and adjusting properties for models like `Aura-1` and `BGE-M3`. Crucially, it also introduces **new capabilities** for several models, including **asynchronous batch processing** for `GPT-OSS-120B` and `GPT-OSS-20B`, **function calling** for `Gemma-Sea-Lion-V4-27B-IT` and `Granite-4.0-H-Micro`, and new **websocket pricing** for `Nova-3`. This **maintenance** and **feature enhancement** update ensures the accuracy and expands the functionality of the **Workers AI platform** by reflecting the latest upstream API changes.

This commit introduces **new Qwen models** to the **Workers AI platform**, expanding the range of available AI capabilities. It adds definition files for the `Qwen3-30b-a3b-fp8` large language model and the `Qwen3-embedding-0.6b` embedding model, making them accessible for use. This **new capability** is accompanied by **documentation updates** in `pricing.mdx` to include their respective costs and an entry in `workers-ai.yaml` for release notes. Users can now leverage these additional Qwen models for their AI applications on Workers AI.

This commit delivers crucial **documentation updates** for the **Workers AI** platform, addressing both feature limits and pricing accuracy. It **updates the maximum number of LoRA adapters** allowed per account from 30 to 100 in `src/content/docs/workers-ai/features/fine-tunes/loras.mdx`. Concurrently, it **corrects the neuron pricing** for the `@cf/deepgram/aura-2-en` and `@cf/deepgram/aura-2-es` models in `src/content/docs/workers-ai/platform/pricing.mdx`, adjusting the cost from 2.73 to 2727.27 neurons per 1k characters input. This **maintenance** ensures users have precise and up-to-date information regarding **Workers AI** capabilities and associated costs.

This commit introduces **new AI models** to the **Workers AI platform**, significantly expanding its capabilities. It adds the **Deepgram Aura 2 English and Spanish text-to-speech models** and the **IBM Granite model**, along with **IBM** as a new model author. The work involves **new capability** additions through model definitions, accompanied by essential **documentation updates** for **pricing information** and **release notes** within the `workers-ai` subsystem. This ensures users have access to the new models and comprehensive, up-to-date information regarding their availability and cost.

This commit **introduces a new AI model**, `@cf/ibm-granite/granite-4.0-h-micro`, to the **Workers AI platform**. It adds the necessary **model definition and schema** in `src/content/workers-ai-models/granite-4.0-h-micro.json`, enabling its use within the platform. Concurrently, the **documentation** in `src/content/docs/workers-ai/platform/pricing.mdx` is updated to include the pricing details for this new capability, ensuring users have complete information. This work expands the available AI models for Workers AI users by adding a new IBM Granite offering.

sigh missed backticks (#25596)