MiniMax: MiniMax M3

Community sentiment

Positive (+68) — 10 positive, 0 negative, 5 neutral across 15 community samples from X and the web, as of 2026-09-03.

Discussion centers on MiniMax M3’s performance and value: users praise its speed, large context window and coding/agent capabilities (including low hallucination rates and effective caching), and many highlight its favorable cost/access options. A minority ask about lineage, language/data coverage, or note limits for some use cases.

• “@jumperz GPT Sol 5.6 is really good at this. Minimax M3 with Roo Code harness in architect mode is also exceptional at this. I would say Minimax M3 + Roo Code is the most comprehensive even with you having to ask.” — X
• “MiniMax-M3 on https://t.co/WM61uG5MDJ: → 512K context → SOTA coding + agent tool use → $0.313/M input, $1.252/M output GPT-4o: $2.50/M. MiniMax-M3: $0.31/M. Save 87%. Same OpenAI SDK. #MiniMax #AI #Coding https://t.co/oBRUMnt0AS” — X
• “@bridgemindai Minimax M3 with no hallucinations damn 🏅” — X
• “am I crazy or Minimax M3 caching is really working well” — X

RuntimeWire coverage

• MiniMax says M3 sent a business email for $0.018
• MiniMax puts four models on GMI Cloud's two-week unlimited tab
• DeepSeek ships V4 Flash with strong agent scores and low API prices
• GLM-5.2 leads July's open-weight LLMs as Inkling tops US models
• DeepReinforce releases Ornith-1.0 for self-scaffolding coding agents
• Z.ai's GLM-5.2 tops open-weight models on Artificial Analysis work benchmark
• RuntimeWire — Weekly Report Week 2 · June 5th – June 11th, 2026
• Kilo Code AI says MiniMax M3 matched Claude Opus 4.8 on a code audit for $0.07