MiniMax M2.7: The $0.30 Model That Matches Claude Opus 4.6

Released: March 18–26, 2026 | Source: Geeky Gadgets

Chinese AI lab MiniMax just dropped M2.7 — a 230-billion-parameter model that benchmarks alongside Claude Opus 4.6 at roughly $0.30 per million tokens. That's not a typo.

What makes M2.7 stand out

  • 230B parameters with 50–100 tokens/second throughput
  • 200,000-token context window for long-document and agentic tasks
  • Runs on older GPU architectures — no H100 required
  • ~$2,000/year operational cost for local deployment
  • Reduces ML workflow manual effort by 30–50%
  • Strong fit for privacy-sensitive and compute-constrained environments

The AI market is bifurcating: ultra-cheap capable models vs. premium frontier models. MiniMax M2.7 is making the case that the gap is closing fast — especially for organizations outside the well-resourced West.

Bottom line: If you're paying Opus prices for coding or agentic workflows, M2.7 is worth a serious look.

Full breakdown at Geeky Gadgets


Subscribe to LGTM for daily AI model news delivered to your inbox.