Skip to content

[Feat] Add Kimi K3.1 Pro and Highspeed models - #3596

Merged
mrubens merged 2 commits into
developfrom
feature/kimi-k31-pro-3akrbfsjzm67c
Oct 27, 2026
Merged

mrubens merged 2 commits into
developfrom
feature/kimi-k31-pro-3akrbfsjzm67c

Conversation

@roomote-roomote

@roomote-roomote roomote-roomote Bot commented Oct 27, 2026 •

Copy link
Copy Markdown
Contributor

​Opened on behalf of @mrubens. Follow up by mentioning @roomote-roomote, in the web UI, or in Slack.

What changed

  • Add Kimi K3.1 Pro to recommended models through Moonshot AI, OpenRouter, Vercel AI Gateway, Kimi for Coding, and OpenCode Go.
  • Add Kimi K3.1 Pro Highspeed through Moonshot AI, Vercel AI Gateway, and Kimi for Coding.
  • Register the two Kimi for Coding models with their context/output limits, reasoning, multimodal input, and tool capabilities in the shared runtime provider configuration.
  • Document full task-model IDs and membership requirements; add regression coverage and a patch changeset.

Why this change was made

The upstream release is named Kimi K3.1 Pro, with a separate Highspeed variant. Exact IDs were verified against Moonshot's API documentation, Kimi for Coding's model documentation, and the live OpenRouter, Vercel, and OpenCode Go catalogs.

Impact

Users can select the new models through five providers (eight task-model routes). Kimi for Coding uses k3.1-pro / k3.1-pro-highspeed on its membership endpoint and resolves metadata through kimi-code-plan-cn.

Validation on the updated branch:

  • 176 targeted types tests and 31 web metadata tests passed, covering the supported provider routes and Kimi runtime/metadata registration.
  • Docs build validation and broken-link checks passed.
  • Pre-push oxlint, residual ESLint, repository typechecks, and knip passed.
  • The existing Requesty provider configuration and metadata implementation exactly match the PR base; no new Requesty support remains in this PR.

Authenticated inference was not exercised. Kimi membership tier controls model and full-context access. The optional pre-commit Judgement evaluation reported incomplete evidence collection/evaluation; that check remains to be completed before merge.

@roomote-community

roomote-community Bot commented Oct 27, 2026 •

Copy link
Copy Markdown
Contributor

No new code issues found. The prior Requesty issue is addressed by removing both new routes and their metadata fallback. See session

Reviewed 5e5b6e5

Comment thread packages/types/src/model-provider-config.ts Outdated
@roomote-roomote

roomote-roomote Bot commented Oct 27, 2026 •

Copy link
Copy Markdown
Contributor Author

Completed in 5e5b6e58: removed both new Requesty Kimi K3.1 routes, the metadata compatibility fallback, related tests, and docs/changeset references. Existing Requesty configuration and metadata implementation match the PR base. The other five providers’ Kimi additions remain; the PR description now reflects eight supported routes. The associated Requesty review thread is resolved by removing the unsupported additions.

Validation: 176 targeted types tests and 31 metadata tests passed; docs build/link validation and pre-push oxlint, residual ESLint, typechecks, and knip passed. No new Requesty additions remain in the net diff. Optional pre-commit Judgement evaluation is still incomplete due to evidence collection/evaluation failure; authenticated inference was not exercised.

@roomote-roomote
roomote-roomote Bot marked this pull request as ready for review October 27, 2026 02:03
@mrubens
mrubens merged commit 62112eb into develop Oct 27, 2026
31 checks passed
@mrubens
mrubens deleted the feature/kimi-k31-pro-3akrbfsjzm67c branch October 27, 2026 02:08
@roomote-roomote roomote-roomote Bot mentioned this pull request Oct 27, 2026
7 of 11 tasks
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