docs: add GPT-6 Astra, Gemini 3/3.8 Flash, Inkling, and Muse Spark 1.2 to model choice

[warp-factories]

Closes QUALITY-2031.

What changed

Audited src/content/docs/agents/inference/model-choice.mdx against the warp-server model catalog (model/types/ai/types.go ConfigurableLlmIds, internal_llm_types.go) and added the following confirmed-shipped models:

• OpenAI: GPT-6 Astra low / medium / high / xhigh / max reasoning levels (gpt-6-astra-*). max is a new reasoning tier beyond the previous "Extra High" ceiling.
• Google: Gemini 3 Flash (gemini-3-flash) and Gemini 3.8 Flash (gemini-3.8-flash).
• Fireworks-hosted: Inkling (inkling-fireworks).
• New "Meta" section: Muse Spark 1.2 (muse-spark-1.2). The catalog shows this model uses ProviderMeta, not Fireworks — it's excluded from FireworksModelIds — so it didn't belong in the Fireworks-hosted table as the request assumed. Gave it its own section instead.

All four additions are present in ConfigurableLlmIds (the list of models configurable via org settings, which is what backs the model_id values this page documents) and have full display-name/context-window/output-token wiring alongside already-documented GA models.

Left out (flagged, not added)

• GPT-6 Astra and Muse Spark 1.2 are both gated by server feature flags (gpt_6_astra, meta_muse_spark in config/features/features.go). I couldn't verify rollout percentage from the repo alone. GPT-6 Astra's inclusion is corroborated by the requester (an engineer) and by production billing logic already wired for it (OpenAI data-residency surcharge list). Muse Spark 1.2 has equivalent code wiring but wasn't explicitly named by the requester — please double check its GA status before merging.
• Several other catalog IDs exist but look like superseded versions of model families already trimmed from this page (e.g. gpt-4o, gpt-4.1, o3/o4-mini, older GPT-5.x variants, claude-4-sonnet, claude-4-opus, claude-4.1-opus, claude-3-5-sonnet/haiku, grok-3, grok-3-mini, grok-4, gemini-3-pro, and older Fireworks entries like GLM 4.6/4.7/5.1, Kimi K2.5, Minimax 2.7, Qwen 3.6/3.7 Plus, GPT-OSS-120B, and the Baseten GLM 4.6 variant). These weren't part of the requested scope and the page already excludes them, so I left them alone rather than guess whether that was intentional deprecation or an existing gap.
• Didn't add "Meta" to the Zero Data Retention provider list in this page, since I have no evidence of Warp's ZDR contractual status with Meta.

Verification

• npm run build — passes.
• trunk check src/content/docs/agents/inference/model-choice.mdx — no new issues (30 pre-existing, unrelated).

[cla-bot]

Thank you for your pull request and welcome to our community. We require contributors to sign our Contributor License Agreement, and we don't seem to have the users @warp-factory on file. In order for us to review and merge your code, each contributor must visit https://cla.warp.dev to read and agree to our CLA. Once you have done so, please comment @cla-bot check to trigger another check.

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Actions	Updated
docs	Ready	Preview	Sep 8, 2026 5:24am UTC

[warp-factories]

This PR was generated with Warp.

Comment @warp-factory on this PR to send it follow-up work.