feat: upgrade MiniMax default model to M3

[octo-patch]

Summary

Upgrade MiniMax model configuration in the minimax provider to include the latest M3 model as the new default.

Changes

• Add MiniMax-M3 to the model selection list and set it as the default (listed first)
• Retain MiniMax-M2.7 and MiniMax-M2.7-highspeed as available alternatives
• Remove deprecated MiniMax-M2.5 / MiniMax-M2.5-highspeed from the list
• Update the capability flags in web-app/src/constants/models.ts (streaming / tool-calls)
• Update the relevant unit test to reference the new default model

Why

MiniMax-M3 is the new flagship model from MiniMax with enhanced reasoning and coding capabilities. Setting it as the default and removing the older M2.5 line keeps the provider aligned with MiniMax's current offering, while M2.7 stays around for users who still want it.

API Reference

• Chat (OpenAI Compatible): https://platform.minimax.io/docs/api-reference/text-openai-api

Testing

• Unit test in web-app/src/lib/__tests__/model-factory.test.ts updated to verify the OpenAI-compatible model factory works with the new MiniMax-M3 model id.
• Existing OpenAI-compatible path in model-factory.ts is unchanged, so all other minimax provider behavior continues to work.

[tokamak-pm]

Review: PR #8251 - feat: upgrade MiniMax default model to M3

Summary

This PR updates the MiniMax provider configuration to add MiniMax-M3 as the new default model, retain M2.7 variants, and remove deprecated M2.5 variants. The corresponding unit test is updated.

Changes Analyzed

1. web-app/src/constants/models.ts

   • Model list changed from [M2.7, M2.7-highspeed, M2.5, M2.5-highspeed] to [M3, M2.7, M2.7-highspeed].
   • supportsStreaming and supportsToolCalls arrays updated consistently.
   • supportsJSON and supportsImages remain empty -- this looks correct if MiniMax-M3 does not advertise those capabilities via the OpenAI-compatible API.

2. web-app/src/constants/providers.ts

   • MiniMax-M3 is added as the first (default) model with description "Latest flagship model with enhanced reasoning and coding."
   • M2.7 description updated from the original flagship description to "Previous flagship model with strong reasoning and coding."
   • M2.7-highspeed retained with the same description.
   • M2.5 and M2.5-highspeed removed entirely.

3. web-app/src/lib/__tests__/model-factory.test.ts

   • Test updated to use MiniMax-M3 instead of MiniMax-M2.7. This validates that the OpenAI-compatible model factory still works with the new model ID.

Observations

• Breaking change for existing users: Users who had MiniMax-M2.5 or MiniMax-M2.5-highspeed selected will no longer see those models in the list. If the model ID was persisted in thread/conversation settings, those threads may fail to use the model. Consider whether a migration or fallback is needed (e.g., auto-switching to M3 if the configured model is no longer available). This is a minor risk -- the UI likely falls back to the first available model, but it is worth verifying.

• No supportsJSON for M3: If MiniMax-M3 supports structured/JSON output (common for newer flagship models), this should be added. Worth checking against the MiniMax API docs.

• PR description is thorough and well-structured. Good API reference link provided.

• Test coverage is adequate for the scope of the change. The test verifies model creation works with the new ID.

• The author (octo-patch) is a new contributor with no prior commits to this repo. The changes are clean and well-reasoned.

Verdict

Straightforward configuration update. The only concern is the impact on users with M2.5 models configured, but that is a product decision (deprecation is intentional per the PR description).

Recommendation: can merge