Skip to content

feat: upgrade MiniMax default model to M3 - #1442

Open
octo-patch wants to merge 1 commit into
NoFxAiOS:devfrom
octo-patch:feature/upgrade-minimax-m2.7
Open

feat: upgrade MiniMax default model to M3#1442
octo-patch wants to merge 1 commit into
NoFxAiOS:devfrom
octo-patch:feature/upgrade-minimax-m2.7

Conversation

@octo-patch

@octo-patch octo-patch commented Mar 29, 2026

Copy link
Copy Markdown

Summary

Upgrade MiniMax default model from M2.5/M2.7 to M3 across the Go backend, frontend constants, and docs. M3 is the latest MiniMax flagship: 512K context window, up to 128K output, and image input support.

M2.7 / M2.7-highspeed remain selectable by users via the existing custom-model field; only the default is changed.

Changes

Backend (Go)

  • mcp/providers.go: DefaultMiniMaxModel MiniMax-M2.7MiniMax-M3
  • mcp/provider/minimax.go: DefaultMiniMaxModel MiniMax-M2.7MiniMax-M3
  • api/handler_ai_model.go: supported-models list default MiniMax-M2.7MiniMax-M3
  • agent/agent.go: resolveModelRuntimeConfig minimax entry — model MiniMax-M2.5MiniMax-M3, URL api.minimax.chatapi.minimax.io (correct endpoint)
  • agent/model_provider_catalog.go: supportedModelProviders minimax DefaultModel MiniMax-M2.5MiniMax-M3
  • store/strategy.go: contextLimitMinimax 1_000_000 (1M)512_000 (512K) to match M3 spec

Frontend (TypeScript)

  • web/src/components/trader/model-constants.ts: AI_PROVIDER_CONFIG.minimax.defaultModel MiniMax-M2.7MiniMax-M3

Docs

  • docs/token-estimation.zh-CN.md: MiniMax row context window 1,000,000 / 800,000512,000 / 409,600

Why M3?

MiniMax-M3 is the latest flagship from MiniMax, with a 512K context window, up to 128K output tokens, and (in OpenAI-compatible mode) image-input support. The OpenAI-compatible API endpoint (https://api.minimax.io/v1) and auth scheme are unchanged, so no client-side wiring needs to be touched.

Compatibility

  • Users who hard-coded MiniMax-M2.7 (or earlier) via the custom-model UI field are unaffected — the change only updates the default model.
  • Older models (MiniMax-M2.5/M2.1/M2/M1) were never on the picker list in dev — only defaults were updated.

Test Plan

  • mcp / agent / api / store compile cleanly (go build ./...)
  • No references to deprecated api.minimax.chat remain in code
  • No 1_000_000 1M context literal remains for minimax
  • Frontend AI_PROVIDER_CONFIG.minimax.defaultModel is MiniMax-M3

@cla-assistant

cla-assistant Bot commented Mar 29, 2026

Copy link
Copy Markdown

CLA assistant check
All committers have signed the CLA.

@cla-assistant

cla-assistant Bot commented Mar 29, 2026

Copy link
Copy Markdown

CLA assistant check
Thank you for your submission! We really appreciate it. Like many open source projects, we ask that you sign our Contributor License Agreement before we can accept your contribution.


PR Bot seems not to be a GitHub user. You need a GitHub account to be able to sign the CLA. If you have already a GitHub account, please add the email address used for this commit to your account.
You have signed the CLA already but the status is still pending? Let us recheck it.

- Add MiniMax-M3 as the new default model
- Update mcp/providers.go and mcp/provider/minimax.go DefaultMiniMaxModel
- Update agent/agent.go fallback defaults (URL + model)
- Update agent/model_provider_catalog.go DefaultModel
- Update api/handler_ai_model.go supported models list
- Update web/src/components/trader/model-constants.ts default
- Update store/strategy.go context limit to 512K (M3 spec)
- Update docs/token-estimation.zh-CN.md context window row
@octo-patch
octo-patch force-pushed the feature/upgrade-minimax-m2.7 branch from bf8af6f to 9f6e8a6 Compare June 7, 2026 05:13
@octo-patch
octo-patch requested a review from SkywalkerJi as a code owner June 7, 2026 05:13
@octo-patch octo-patch changed the title feat: upgrade MiniMax default model to M2.7 and add to README docs feat: upgrade MiniMax default model to M3 Jun 7, 2026
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