feat: add MiniMax as LLM judge provider (default model: MiniMax-M3)#1263
Open
octo-patch wants to merge 2 commits into
Open
feat: add MiniMax as LLM judge provider (default model: MiniMax-M3)#1263octo-patch wants to merge 2 commits into
octo-patch wants to merge 2 commits into
Conversation
Add MiniMax (https://www.minimax.io) as a first-class LLM judge provider, enabling evaluation tasks to use MiniMax models (M2.7, M2.5, M2.5-highspeed) via the existing ProviderFactory. Changes: - MiniMaxProvider (sync) and AsyncMiniMaxProvider extending ServerInterface and AsyncServerInterface respectively - OpenAI-compatible API via base_url routing to api.minimax.io/v1 - Temperature clamping to MiniMax's [0.0, 1.0] range - Automatic <think>...</think> tag stripping for reasoning models - Factory registration: 'minimax' and 'async_minimax' api_type values - 31 unit tests + 3 integration tests Usage: export API_TYPE=minimax export MINIMAX_API_KEY=your-key
Contributor
|
why cant we use openai format api to request minimax model? |
kcz358
reviewed
Apr 12, 2026
Comment on lines
+55
to
+66
| try: | ||
| from openai import OpenAI | ||
|
|
||
| self.client = OpenAI( | ||
| api_key=self.api_key, | ||
| base_url=self.MINIMAX_BASE_URL, | ||
| ) | ||
| self.use_client = True | ||
| except ImportError: | ||
| eval_logger.warning( | ||
| "OpenAI client not available, falling back to requests for MiniMax" | ||
| ) |
Collaborator
There was a problem hiding this comment.
Using OpenAI as the base request caller. I agree that this could be similar as openai provider
- Update MiniMaxProvider docstring to mention MiniMax-M3 (default; 512K context, 128K max output, image input supported) - Keep MiniMax-M2.7 and MiniMax-M2.7-highspeed as supported models - Remove references to older M2.5 / M2.5-highspeed models - Update unit and integration tests to default to MiniMax-M3
octo-patch
changed the title
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
3 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
minimaxandasync_minimaxapi_type), enabling evaluation tasks to leverage MiniMax models via the existingProviderFactoryMiniMax-M2.7andMiniMax-M2.7-highspeedremain as supported alternativesMiniMaxProvider) and async (AsyncMiniMaxProvider) implementations follow the same patterns as the existing OpenAI providers[0.0, 1.0]range and automatic<think>...</think>tag stripping for reasoning modelsFiles changed (5 files)
lmms_eval/llm_judge/providers/minimax.pyServerInterfacelmms_eval/llm_judge/providers/async_minimax.pyAsyncServerInterfacelmms_eval/llm_judge/providers/__init__.pylmms_eval/llm_judge/factory.pyminimaxandasync_minimaxinProviderFactorytest/eval/test_minimax_provider.pySupported models
MiniMax-M3MiniMax-M2.7MiniMax-M2.7-highspeedUsage
Or programmatically:
Test plan
MiniMax-M3)MINIMAX_API_KEY)