MiniMax's M3 model beats GPT-5.5 and Gemini 3.1 Pro on coding benchmarks
1M token context, 5x faster processing, and 20x less compute…
Deep Dive
Shanghai-based AI startup MiniMax launched M3, a coding-focused model that can process up to 1 million tokens at once—five times more than its predecessor. It reduces computational requirements to as little as one-twentieth of previous levels, slashing inference costs, and outperformed OpenAI’s GPT-5.5 and Google’s Gemini 3.1 Pro on the SWE-Bench Pro coding benchmark for handling long, complex programming tasks.
Key Points
- Designed for long and complex coding tasks with 1M token context, processing 5x faster than M2.7
- Reduces computational requirements to 1/20th of previous levels, slashing inference costs
- Outperforms GPT-5.5 and Gemini 3.1 Pro on SWE-Bench Pro benchmark
Why It Matters
MiniMax M3 could disrupt enterprise coding agents with superior performance at lower cost.