# GPT-6 Astra and Claude Fable 5.1 are now supported in Grain > Choose Astra through OpenAI or a Codex-backed Executor, and bring Fable 5.1 to ambitious work through Grain’s Claude execution paths. - Canonical page: [/blog/gpt-6-astra-and-claude-fable-5-1-in-grain](https://grain.sh/blog/gpt-6-astra-and-claude-fable-5-1-in-grain) - Agent-readable page: [/agents/blog--gpt-6-astra-and-claude-fable-5-1-in-grain.md](https://grain.sh/agents/blog--gpt-6-astra-and-claude-fable-5-1-in-grain.md) - Section: Blog - Category: Models - Published: September 4, 2026 - Reading time: 5 min read ## Two frontier models, one place to put them to work The model landscape moves quickly. The useful question for a team is not only what is new, but whether the model can enter the same workflow as the repositories, documents, machines, and review rules the team already uses. Grain now supports GPT-6 Astra and Claude Fable 5.1. Both appear through the model-selection flow, so you can choose the right engine for a demanding task without creating a separate process around it. Availability still follows the access on the OpenAI or Claude account you connect. ## Astra through OpenAI and Codex GPT-6 Astra can be selected through a connected OpenAI path or on a Codex-backed Grain Executor. It is also available in per-teammate model settings, which lets a specialist use Astra without changing the default for the whole workspace. Astra is designed for difficult end-to-end work across reasoning, coding, computer use, research, and document creation. In Grain, that capability sits behind the same task brief, machine selection, isolated branch, approval policy, and review surface as the rest of the model roster. - Use a connected OpenAI account where Astra access is enabled. - Run it through Codex on a connected computer with Grain Executor. - Assign it to one AI teammate or choose it for one task. - Keep the work inside Grain’s existing repository and review controls. ## Fable 5.1 for ambitious, long-running work Claude Fable 5.1 is available through Grain’s Claude execution paths for accounts that have access. It is a natural choice for broad coding changes, document-heavy analysis, and work that needs to keep a plan intact across many steps. You can use Fable directly or place it in a pipeline: let it map the work, send implementation to another model, and return the result for verification. Grain preserves the task, artifacts, and execution trail while each model does the part it is best suited to handle. ## Choice without a new control panel These models arrive in the shared picker used by task creation and ongoing conversations. Online computers appear first, unavailable choices explain what is missing, and a selected Executor model keeps the machine that can run it attached to the task. That consistency matters more as the roster grows. A model choice should travel with the work, not become another setting a teammate has to reconstruct every time the conversation moves. ## Start with the job Use Astra when the work benefits from OpenAI’s strongest end-to-end reasoning and tool use. Use Fable 5.1 when a long, ambitious coding or knowledge-work task needs sustained planning. Or leave the choice on Auto and let Grain place routine work for you. Open the model picker in a new task or task conversation to see the models available across your connected accounts and computers. ## About Grain Grain is a multi-agent work platform. A person describes an outcome as a Mission, Grain coordinates the models, tools, files, repositories, and connected computers needed to do the work, and the result returns for review. Start with [the complete agent-readable index](https://grain.sh/agents/index.md) or read [the full site corpus](https://grain.sh/llms-full.txt).