Add direct MiniMax-M3 support for minimax and minimax-cn

[XcluEzy7]

Add direct MiniMax-M3 support for minimax and minimax-cn

MiniMax released M3, but the registry only has M2.7 and M2.7-highspeed on the minimax and minimax-cn providers, so M3 isn't addressable as a direct model today. I'd like to add it.

What

• Add MiniMax-M3 to minimax and minimax-cn in packages/ai/src/models.generated.ts — anthropic-messages API, text+image input, reasoning, 512K context, 128K max tokens. Base URL matches the existing provider (api.minimax.io/anthropic and api.minimaxi.com/anthropic).
• Update scripts/generate-models.ts so minimaxDirectSupportedIds includes M3 and the 512K context applies only to M3; M2.7 and M2.7-highspeed keep 204800 / 131072.
• Add a unit test under packages/ai/test/ covering baseUrl, api, contextWindow, maxTokens, reasoning, and input modalities for both providers.
• Document M3 and MINIMAX_CN_API_KEY in packages/ai/README.md.

Notes

• No new env vars. MINIMAX_API_KEY and MINIMAX_CN_API_KEY already cover auth (verified in env-api-keys.ts).
• No CHANGELOG.md edit per CONTRIBUTING.
• Cost numbers I plan to use (input 0.6, output 2.4, cacheRead 0.12, cacheWrite 0) are not yet sourced from a pricing doc — happy to revise if maintainers have authoritative numbers.

Code is staged on feature/minimax-m3 on my fork. I'll open the PR after lgtm.

[HatemMn]

Any update on this ?

I really need this model