Skip to content

chore(aihubmix): sync September model catalog - #6268

Open
d2bz wants to merge 1 commit into
anomalyco:devfrom
d2bz:chore/aihubmix-september-catalog
Open

chore(aihubmix): sync September model catalog#6268
d2bz wants to merge 1 commit into
anomalyco:devfrom
d2bz:chore/aihubmix-september-catalog

Conversation

@d2bz

@d2bz d2bz commented Sep 4, 2026

Copy link
Copy Markdown

Summary

Sync the AIHubMix provider catalog with three current service entries:

  • Add claude-fable-5-1
  • Add gemini-3.8-flash
  • Correct pricing and max output for doubao-seed-2-0-mini-260428

Changes

  • Scope is limited to providers/aihubmix/models.
  • Inherit upstream metadata through base_model.
  • Add AIHubMix pricing, cache pricing, service limits, modalities, and reasoning controls.
  • Override Gemini input modalities to the service-supported text, image, audio, and video set.
  • Correct Doubao input/output pricing from 0.03/0.28 to 0.0282/0.282 USD/MTok.
  • Correct Doubao max output from 128,000 to 32,000 tokens.

AIHubMix source facts

Facts were retrieved with AIHubMix Models API validator 2026.08.31.3. Prices are USD per 1M tokens and limits are tokens.

Model Input / output / cache-read prices Context / max output Input modalities
claude-fable-5-1 11 / 55 / 0.275 (cache write: 13.75) 1,000,000 / 128,000 text, image
gemini-3.8-flash 0.75 / 3.75 / 0.075 1,048,576 / 65,536 text, image, audio, video
doubao-seed-2-0-mini-260428 0.0282 / 0.282 / 0.00564 256,000 / 32,000 text, image, video

Validation

  • bun run validate
  • bun run build in packages/web
  • git diff --check
  • AIHubMix Models API validator: 3/3 static configurations passed
  • Paid live validation: all configured reasoning levels, JSON schema, and image cases returned HTTP 200
  • Generated local _api.json contained all three AIHubMix entries with expected facts
  • OpenCode 1.18.20 loaded the local combined registry and listed all three models
  • Actual OpenCode calls returned the expected unique markers for all three models
  • OpenCode TUI manual test switched in-session from Claude Fable 5.1 to Gemini 3.8 Flash and both calls succeeded

@github-actions

github-actions Bot commented Sep 4, 2026

Copy link
Copy Markdown
Contributor

Action items

  • [high] [violation] providers/aihubmix/models/claude-fable-5-1.toml:3 - Check: Every toggle must have a leading top-of-file wire-path comment. Why: reasoning_options includes { type = "toggle" } with no leading comment documenting the AIHubMix control (e.g. Messages $.thinking.type = enabled|disabled|adaptive). Sync strips mid-file comments, so the wire path is lost and the entry breaks the reasoning-options policy. Action: Add a leading header comment above the first key that states the exact toggle (and effort) wire fields for this host, matching peers such as claude-opus-5.toml.
  • [high] [violation] providers/aihubmix/models/claude-fable-5-1.toml:11 - Check: After base_model, provider files must be override-only — do not restate identical lab values. Why: [limit] sets context = 1_000_000 and output = 128_000, which already match models/anthropic/claude-fable-5-1.toml, so the block is a redundant copy rather than a real host delta. Action: Remove the identical [limit] section and inherit limits from base_model unless AIHubMix actually differs.

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