Skip to content

feat(catalog): add the Claude 5.5 / GPT-6 Sol / Gemini 4 Argon generation - #16

Merged
dorukozgen merged 1 commit into
mainfrom
feat/catalog-oct-2026-refresh
Oct 22, 2026
Merged

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

Conversation

@dorukozgen

Copy link
Copy Markdown
Contributor

Catalog pass for everything that shipped since the 2026-09-21 refresh, sourced from each provider's own documentation (and, for OpenRouter and NVIDIA NIM, from their live model listings; for Codex and Claude Code, from their bundled model lists). Additions only: no model id is removed or renamed. 184 → 237 models.

Added

Provider New
anthropic, claude-code Claude Opus 5.5, Sonnet 5.5, Haiku 5.5
openai GPT-6 Sol, GPT-6 Luna
codex-cli GPT-6.1 Sol, GPT-6 Sol, GPT-6 Luna
gemini-api Gemini 4 Argon
antigravity-sdk Gemini 4 Argon High / Medium / Low
grok Grok 4.7
mistral Mistral Large 4 (mistral-large-4-0, public preview)
openrouter Opus 5.5, Sonnet 5.5, Haiku 5.5, GPT-6.1 Sol, GPT-6 Sol, GPT-6 Luna, Gemini 4 Argon, Grok 4.7, DeepSeek V4.1 Flash 1017, GLM 5.3 Prime, GLM 5.3 FlashX
github-copilot Opus 5.5, Sonnet 5.5, Haiku 5.5, GPT-6.1 Sol, GPT-6 Sol, GPT-6 Luna, Gemini 4 Argon, Grok 4.7
perplexity Opus 5.5, Sonnet 5.5, Haiku 5.5, Fable 5.1, GPT-6.1 Sol, GPT-6 Sol, GPT-6 Luna, GPT-6 Astra, Gemini 4 Argon, Grok 4.7, Nemotron 3.5 Super, DeepSeek V4 Pro 0813, GLM 5.3 Flash
nvidia-nim Nemotron 3.5 Super VL, DeepSeek V4.1 Flash, Step 5 Preview
cloudflare-workers-ai DeepSeek V4.1 Flash 1017

Notes on the new entries:

  • Claude Sonnet 5.5 cache reads are $0.10 (Anthropic cut them from $0.20 on 2026-10-07); Opus 5.5 cache reads are $0.20 (0.05x input). Haiku 5.5 records the ≤100K-token rate card; its note carries the $0.50 / $2.50 rate for longer prompts.
  • Claude Code resolves opus / sonnet / haiku to the 5.5 models from v2.1.280 / v2.1.284 / v2.1.293. All three run with a native 1M window, so there are no [1m] twins, and Claude Code defaults their effort to medium.
  • Gemini 4 Argon records its introductory price ($2 / $10 through 2026-11-30, $4 / $20 from 2026-12-01). It ignores sampling parameters, so temperature = false.
  • Codex defaults GPT-6.1 Sol to low effort (its ultra sends xhigh plus delegation); GPT-6 Luna has no ultra tier.

Defaults

Defaults follow each vendor's own default:

  • anthropic, claude-code: Opus 5 → Opus 5.5 (Claude Code's default model on every plan since v2.1.280)
  • codex-cli: GPT-5.6 Sol → GPT-6.1 Sol (Codex's default since CLI 0.159.1). GPT-6 Astra moves up to match Codex's picker order.

Every other default is unchanged. Gemini 4 Argon is paid-tier only, so gemini-api stays on 3.1 Pro Preview.

Other changes

  • deepseek-flash last_updated → 2026-10-17: the same id now serves the V4.1-Flash-1017 snapshot.

Deliberately not done

  • GPT-6.1 Sol is not on the openai provider. Like GPT-6 Astra, its tool calling requires the Responses API. GPT-6 Sol and Luna support Chat Completions function calling on the same terms as GPT-5.4 and later.
  • Antigravity Claude Opus / Sonnet 5.5: only the Enterprise slugs (claude-{opus,sonnet}-5-5-{low,medium,high,max}) could be verified; the consumer-plan slug could not.
  • Kimi Code K2.8 Preview is served only under the subscription id kimi-for-coding, which does not fit the catalog's kimi-k* ids.
  • antigravity-cli now supports --model; whether to list explicit models depends on the app's bridge.
  • Cloudflare PhoneLLM Mini (OpenAI-compatible endpoint unverified) and two access-request models; OpenRouter models from vendors the catalog does not track there (Mistral Large 4, Qwen 3.8 Max, Muse, MiMo, …).

Unverified id: GitHub has not published the Copilot CLI id for Gemini 4 Argon. gemini-4-argon follows the other Copilot Gemini ids, which match Google's API ids.

Follow-ups (not in this PR)

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

  • nvidia-nim: the default nemotron-3-super-120b-a12b was deprecated on 2026-10-19; kimi-k3 and mistral-nemotron are gone from /v1/models; glm-5.3 is unsupported after 2026-10-28.
  • perplexity: Sonar API support ended on 2026-09-27 (sonar-pro is the default); openai/gpt-5.4, gpt-5.2, gpt-5.1 and gpt-5-mini stop being accepted on 2026-10-24; 10 more ids were dropped from the model list in October.
  • github-copilot: 8 catalog ids retired between 2026-09-01 and 2026-10-19; claude-opus-5, claude-opus-4.8 and gpt-5.6-terra retire on 2026-11-16.
  • gemini-api: Gemini 3.7 Flash and 3.5 Flash were deprecated on 2026-10-08 (routed to 3.8 / 3.6 Flash).
  • antigravity-sdk: Claude 4.6 and GPT-OSS 120B are removed on 2026-11-02; Gemini 3.1 Pro (the default) is now Enterprise-only.
  • Drift: OpenRouter gpt-5.6-sol $2 / $10 → $4 / $20; Perplexity gpt-5.6-sol $5 / $30 → $4 / $20; Kimi K3 effort default is now max and it bills a $3 cache write; Gemini 3.6 / 3.7 / 3.8 Flash and 3.5 Flash-Lite ignore temperature; the claude-code claude-opus-5 200K override and claude-opus-5[1m] are redundant now that Opus 4.7+ runs 1M natively.

Verification

bun run validate   ✓ 17 providers, 237 models
bun run build      ✓ dist/ regenerated
bun test           ✓ 31 pass, 0 fail

…tion

Add the models that shipped since the 2026-09-21 catalog pass, sourced from
each provider's own documentation and live model listings. Additions only:
no model id is removed or renamed. 184 -> 237 models.

Added:
- anthropic, claude-code: Claude Opus 5.5, Sonnet 5.5, Haiku 5.5
- openai: GPT-6 Sol, GPT-6 Luna
- codex-cli: GPT-6.1 Sol, GPT-6 Sol, GPT-6 Luna
- gemini-api: Gemini 4 Argon
- antigravity-sdk: Gemini 4 Argon high/medium/low
- grok: Grok 4.7
- mistral: Mistral Large 4 (public preview)
- openrouter: Opus 5.5, Sonnet 5.5, Haiku 5.5, GPT-6.1 Sol, GPT-6 Sol,
  GPT-6 Luna, Gemini 4 Argon, Grok 4.7, DeepSeek V4.1 Flash 1017,
  GLM 5.3 Prime, GLM 5.3 FlashX
- github-copilot: Opus 5.5, Sonnet 5.5, Haiku 5.5, GPT-6.1 Sol, GPT-6 Sol,
  GPT-6 Luna, Gemini 4 Argon, Grok 4.7
- perplexity: Opus 5.5, Sonnet 5.5, Haiku 5.5, Fable 5.1, GPT-6.1 Sol,
  GPT-6 Sol, GPT-6 Luna, GPT-6 Astra, Gemini 4 Argon, Grok 4.7,
  Nemotron 3.5 Super, DeepSeek V4 Pro 0813, GLM 5.3 Flash
- nvidia-nim: Nemotron 3.5 Super VL, DeepSeek V4.1 Flash, Step 5 Preview
- cloudflare-workers-ai: DeepSeek V4.1 Flash 1017

Defaults follow each vendor's own default: anthropic and claude-code move to
Opus 5.5, codex-cli to GPT-6.1 Sol (Codex's default since CLI 0.159.1), and
GPT-6 Astra moves up to match Codex's picker order. Every other default is
unchanged. deepseek-flash's last_updated moves to 2026-10-17, when the same
id switched to the V4.1-Flash-1017 snapshot.

GPT-6.1 Sol is held back from the `openai` provider because, like GPT-6
Astra, its tool calling requires the Responses API.
@dorukozgen
dorukozgen merged commit 22aa21e into main Oct 22, 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