community pool / model answer

MiniMax-M2.7

chat · MiniMaxAI/MiniMax-M2.7

32.2 tok/s · model median · 34 runs in the model record.

Reference rigQuantizationMedian resultBasisTTFT (ms)Peak VRAMMax contextEngines
Arc Pro B70 32GB ×44-bit37.7 tok/smeasured-29.2 GB24,576llama.cpp, vllm
GB10 Grace Blackwell 128GB4-bit30.9 tok/sreported807117 GB108,000llama.cpp
Radeon AI Pro R9700 32GB ×34-bit25.2 tok/smeasured771- GB32,768llama.cpp

Context tested

The table above is the tested context for each exact rig and quantization cell. Empty fields remain empty rather than being filled with an estimate.