docs: update models for GPT-6 Astra support

[tembo]

What changed

• Added GPT-6 Astra (gpt-6-astra) to the supported models table in features/models.mdx.
• Added GPT-6 Astra to the Codex, OpenCode, Pi, and fx model lists in features/agents.mdx, and updated the Codex description to note it also supports max/ultra reasoning effort.
• Added a GPT-6 Astra row to the Tembo Gateway pricing table in resources/pricing.mdx.
• Added GPT-6 Astra to the Codex model list in resources/snippets/coding-agents-table.mdx.

Why

The monorepo added runtime support for GPT-6 Astra across the Codex, OpenCode, Pi, and fx harnesses, but the docs only listed models through GPT-5.6. Without this update, the docs undercount the supported model catalog and omit gateway pricing for the new model.

Evidence

• Merged PR: tembo/monorepo#11044 "feat(models): add GPT-6 Astra runtime support" (commit c62bac4793), which adds gpt-6-astra to apps/api/src/server/models/index.ts (requiresApiKey: true, reasoningLevels: ['low','medium','high','xhigh','max']) and to the Codex/OpenCode/Pi/fx harness definitions under packages/agent-runtime/src/runtime/harness/*/definition.ts. Codex's reasoningLevelsByModel additionally grants max and ultra for gpt-6-astra, matching GPT-5.6 Sol/Terra.
• Test coverage in apps/api/tests/models.test.ts (exposes GPT-6 Astra on every compatible harness) confirms it is exposed on codex/opencode/pi/fx and ordered before gpt-5.6-sol in each harness's model list.
• Pricing: apps/litellm/litellm-config.json defines gpt-6-astra costs (input_cost_per_token: 0.00001, output_cost_per_token: 0.00005, cache_creation_input_token_cost: 0.0000125, cache_read_input_token_cost: 0.000001). The existing gateway margin constant CREDIT_MARGIN = 1.15 in apps/api/src/services/tokenUsage.ts applied to these base costs reproduces the documented per-1M-token prices ($11.50 / $57.50 / $14.375 / $1.15) — verified against the existing kimi-k3 row, whose displayed price is exactly its litellm-config.json cost × 1.15.

Confidence

This is a merged, fully tested feature (dedicated model + harness + pricing tests). All new model list entries and the pricing row are placed in the same position as the source code's model ordering, and the pricing math is validated against an existing, unambiguous example (kimi-k3) rather than guessed.

cc @benja

Want tembo to make any changes? Add a comment with @tembo and i'll get back to work!

      

---

Note

Low Risk
Documentation-only updates to model lists and pricing; no application code or runtime behavior changes.

Overview
Documents GPT-6 Astra (gpt-6-astra) now that runtime support exists in Codex, OpenCode, Pi, and fx.

The supported models table in features/models.mdx adds the model as OpenAI with BYOK required. Tembo Gateway pricing in resources/pricing.mdx adds a row ($11.50 / $57.50 input/output per 1M tokens, with cache create/read).

Agents docs expand the Codex blurb to cover the GPT-6 family and note that GPT-6 Astra supports max and ultra reasoning (alongside GPT-5.6 Sol/Terra). GPT-6 Astra is prepended to the model lists for Codex, OpenCode, Pi, and fx. The reusable coding agents snippet lists gpt-6-astra among Codex’s additional models.

^{Reviewed by Cursor Bugbot for commit d8623d8. Bugbot is set up for automated code reviews on this repo. Configure here.}

[mintlify]

Preview deployment for your docs. Learn more about Mintlify Previews.

Project	Status	Preview	Updated
tembo	🟢 Ready	View Preview	Sep 5, 2026, 1:06 PM

💡 Tip: Enable Automations to automatically generate PRs for you.

[tembo]

Requesting review from @DarrenBaldwin07 who has experience with the following files modified in this PR:

• features/models.mdx
• resources/pricing.mdx