Minimax M2.7 230B-A10B port#22

Open

joshue031 wants to merge 1 commit intodanveloper:mainfrom

joshue031:minimax-m2.7-metal-port

joshue031 commented Apr 18, 2026

Adapts the codebase to run MiniMax M2.7

current status: local inference works, tool calling works
runs at ~2-3 tokens/s on a Macbook Pro M4 Pro (48GB RAM)
known limitation: prefill is too slow (also 2-3 tokens/s) for practical long-context agent workflows
made using Codex/GPT-5.4


          Minimax M2.7 port; made with Codex/GPT-5.4

7ced576

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet