Skip to content

feat(catalog): add Claude Fable 5.2, GPT-6.1 Terra and Gemini 4 Carbon/Hydrogen - #17

Merged
dorukozgen merged 1 commit into
mainfrom
feat/catalog-nov-2026-refresh
Nov 15, 2026
Merged

dorukozgen merged 1 commit into
mainfrom
feat/catalog-nov-2026-refresh

Conversation

@dorukozgen

Copy link
Copy Markdown
Contributor

Catalog pass for everything that shipped since the 2026-10-22 refresh, sourced from each provider's own documentation (and, for OpenRouter and NVIDIA NIM, from their live model listings; for Codex, from its bundled models.json; for GitHub Copilot, from the github/docs model tables). Additions only: no model id is removed or renamed and no default changes. 237 → 269 models.

Added

Provider New
anthropic, claude-code Claude Fable 5.2
codex-cli GPT-6.1 Terra
gemini-api Gemini 4 Carbon, Gemini 4 Hydrogen
antigravity-sdk Gemini 4 Carbon High / Medium / Low, Gemini 4 Hydrogen XHigh / High / Medium / Low
kimi-api Kimi K3.1 Pro, Kimi K3.1 Pro Highspeed
kimi-code Kimi K3.1 Pro
mistral GLM 5.3 Flash (zai-glm-5-3-flash), DeepSeek V4.1 Flash 1017 (deepseek-v4-1-flash-1017, public preview)
nvidia-nim MiniMax M3.1 Flash, Mistral Large 4 (mistral-large-4-1t-a52b)
cloudflare-workers-ai Inkling 2 Small
openrouter Fable 5.2, GPT-6.1 Terra, Gemini 4 Carbon, Gemini 4 Hydrogen, Kimi K3.1 Pro
github-copilot Fable 5.2, GPT-6.1 Terra, Gemini 4 Carbon, Gemini 4 Hydrogen
perplexity Fable 5.2, Gemini 4 Carbon, Gemini 4 Hydrogen

Notes on the new entries:

  • Claude Fable 5.2 (2026-11-09) is $8 / $40 with $0.20 cache reads (0.025x input). Claude Code resolves fable to it from v2.1.321 and defaults it to medium effort (the API default is high); from the same version best follows opus.
  • GPT-6.1 Terra (2026-10-26) is $1 / $5. Codex bundles it from CLI 0.165.0 and defaults it to low effort.
  • Gemini 4 Carbon (GA 2026-11-10) is paid-tier only, at introductory pricing ($2 / $10 through 2026-11-30, $4 / $20 from 2026-12-01). Gemini 4 Hydrogen (GA 2026-11-12) is $1.50 / $7.50, is on the free tier, and adds an xhigh thinking level (default high). Both ignore sampling parameters, so temperature = false. Antigravity added both to agy for GEMINI_API_KEY users in CLI 1.5.8.
  • Kimi K3.1 Pro (2026-10-23) is $2 / $10. The API defaults its effort to max, Kimi Code to high. Highspeed is the same model at twice the price.
  • NIM serves MiniMax M3.1 Flash and Mistral Large 4 with a 262,144-token window, and their NIM request schemas document no response_format, so structured_output is left unset.

Picker order

To make room, nine existing priorities shift by one or two; relative order is unchanged:

  • github-copilot: auto 230 → 231, claude-opus-5.5 229 → 230, gpt-6-luna 155 → 154, gpt-5.6-terra 154 → 153
  • perplexity: claude-fable-5-1 187 → 186, claude-fable-5 186 → 185, claude-opus-4-8 185 → 184, gemini-4-argon 79 → 77, gemini-3.8-flash 78 → 76

Deliberately not done

  • GPT-6.1 Terra is not on the openai provider. Like GPT-6.1 Sol and GPT-6 Astra, its tool calling requires the Responses API.
  • Claude Mythos 5.2 (verified organizations only); OpenRouter -pro variants (the same models with reasoning.mode = pro); GPT-6 Rosalind and GPT-6 Cyber (gated access); Gemini Omni (video generation) and Gemini Decision 1 (Decisions API only); GLM 5.3 FlashX Legal (legal-domain variant); Nemotron 3.5 Bolt E4B (edge model); the Antigravity Orchestrator slugs (API-key support unverified); Perplexity's openai/gpt-6.1-terra (used internally but not on its model list); Kimi K3.1 Pro Highspeed on kimi-code (K2.7 Code HighSpeed is not listed there either).

Unverified ids: GitHub has not published the Copilot CLI ids for Fable 5.2, Gemini 4 Carbon and Gemini 4 Hydrogen. claude-fable-5.2, gemini-4-carbon and gemini-4-hydrogen follow the other Copilot Claude and Gemini ids. Mistral does not document the default effort for GLM 5.3 Flash; high is used.

Follow-ups (separate PR)

Retirements and drift found during the research. They are left out because removals are breaking and some need a new default picked:

  • github-copilot: claude-opus-5, claude-opus-4.8 and gpt-5.6-terra retire on 2026-11-16; eight catalog ids are already retired; six more retire on 2026-12-14.
  • cloudflare-workers-ai: 12 catalog models return 410 Gone from 2026-11-20.
  • nvidia-nim: the default nemotron-3-super-120b-a12b was deprecated on 2026-10-19; kimi-k3, mistral-nemotron and glm-5.3 are gone from /v1/models.
  • antigravity-sdk: Claude 4.6 and GPT-OSS 120B are removed on 2026-11-18; the default Gemini 3.1 Pro is Enterprise-only.
  • perplexity: GPT-5.4, 5.2, 5.1 and 5 Mini are no longer accepted, and more ids are off the model list.
  • mistral: Large 4 is GA at $0.70 / $2.10, takes low / medium / high effort (no longer none), and mistral-large-latest now points at it.
  • kimi-api: the default kimi-k2.6 is closed to accounts created after 2026-10-31.

Verification

bun run validate   ✓ 17 providers, 269 models
bun run build      ✓ dist/ regenerated
bun test           ✓ 31 pass, 0 fail

…n/Hydrogen

Add the models that shipped since the 2026-10-22 catalog pass, sourced from
each provider's own documentation and live model listings. Additions only:
no model id is removed or renamed and no default changes. 237 -> 269 models.

Added:
- anthropic, claude-code: Claude Fable 5.2
- codex-cli: GPT-6.1 Terra
- gemini-api: Gemini 4 Carbon, Gemini 4 Hydrogen
- antigravity-sdk: Gemini 4 Carbon high/medium/low, Gemini 4 Hydrogen
  xhigh/high/medium/low
- kimi-api: Kimi K3.1 Pro, Kimi K3.1 Pro Highspeed
- kimi-code: Kimi K3.1 Pro
- mistral: GLM 5.3 Flash, DeepSeek V4.1 Flash 1017 (public preview)
- nvidia-nim: MiniMax M3.1 Flash, Mistral Large 4
- cloudflare-workers-ai: Inkling 2 Small
- openrouter: Fable 5.2, GPT-6.1 Terra, Gemini 4 Carbon, Gemini 4 Hydrogen,
  Kimi K3.1 Pro
- github-copilot: Fable 5.2, GPT-6.1 Terra, Gemini 4 Carbon, Gemini 4 Hydrogen
- perplexity: Fable 5.2, Gemini 4 Carbon, Gemini 4 Hydrogen

GPT-6.1 Terra is held back from the `openai` provider because, like GPT-6.1
Sol, its tool calling requires the Responses API.

To make room in the picker order, nine github-copilot and perplexity
priorities shift by one or two; relative order is unchanged. The claude-code
comments now note that `fable` resolves to Fable 5.2 and `best` to Opus 5.5
from v2.1.321.
@dorukozgen
dorukozgen merged commit 6dbcbe2 into main Nov 15, 2026
2 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant