KG narrative

[KG] MiniMax — momentum

What the brain wrote

MiniMax, the Shanghai-based foundation-model startup founded by ex-SenseTime executives, just shipped M3, an LLM that exceeds human gold-medal performance on math benchmarks via its MaxProof technique. That’s no lab stunt — M3 scored 55 on the AI Index, positioning it as an open-source leader. The company isn’t stopping there: it also released M2.5, MMX-CLI, MaxHermes, and Music 2.6. But MiniMax’s ambition is broader. It competes directly with Google, OpenAI, and Anthropic, and specifically targets Claude Code — a product it also uses. Partnerships with Nous Research and Together AI suggest a play for open-source ecosystem leverage rather than pure proprietary moat. With SambaNova hitting 850 tokens/second on M2.7 via hybrid inferencing, deployment velocity is real. The open question: can MiniMax sustain this multi-model offensive without the revenue base of its US rivals?

Knowledge-graph narrative

Entity

MiniMax

Angle

momentum

Key points

•M3 exceeds human gold-medal math performance via MaxProof technique
•Scored 55 on AI Index, signaling open-source leadership
•Direct competitor to OpenAI, Anthropic, Google, and Claude Code
•Partners with Nous Research and Together AI for ecosystem leverage
•SambaNova achieves 850 t/s on M2.7 via hybrid H200-RDU inferencing

Raw payload

{
  "entity_slug": "minimax",
  "entity_name": "MiniMax",
  "entity_type": "company",
  "title": "MiniMax M3 Breaks Math Ceiling, Takes on OpenAI and Anthropic",
  "narrative": "MiniMax, the Shanghai-based foundation-model startup founded by ex-SenseTime executives, just shipped M3, an LLM that exceeds human gold-medal performance on math benchmarks via its MaxProof technique. That’s no lab stunt — M3 scored 55 on the AI Index, positioning it as an open-source leader. The company isn’t stopping there: it also released M2.5, MMX-CLI, MaxHermes, and Music 2.6. But MiniMax’s ambition is broader. It competes directly with Google, OpenAI, and Anthropic, and specifically targets Claude Code — a product it also uses. Partnerships with Nous Research and Together AI suggest a play for open-source ecosystem leverage rather than pure proprietary moat. With SambaNova hitting 850 tokens/second on M2.7 via hybrid inferencing, deployment velocity is real. The open question: can MiniMax sustain this multi-model offensive without the revenue base of its US rivals?",
  "key_points": [
    "M3 exceeds human gold-medal math performance via MaxProof technique",
    "Scored 55 on AI Index, signaling open-source leadership",
    "Direct competitor to OpenAI, Anthropic, Google, and Claude Code",
    "Partners with Nous Research and Together AI for ecosystem leverage",
    "SambaNova achieves 850 t/s on M2.7 via hybrid H200-RDU inferencing"
  ],
  "angle": "momentum",
  "neighborhood_size": 17,
  "generated_at": "2026-07-29T15:49:35.895049+00:00"
}