MiniMax launches M3.1 Flash, a 428B‑parameter MoE model with 1M‑token context

MiniMax released the M3.1 Flash model on Monday, offering a 428 billion‑parameter mixture‑of‑experts architecture with about 23 billion active parameters and a 1 million‑token context window. API pricing is $0.15 per 1 M input tokens and $0.60 per 1 M output tokens, and the model is free on AI Gateway until Oct 22. The model scores 42 on the Artificial Analysis Intelligence Index and 73.7% on the DeepSWE v1.1 coding benchmark, matching Deepseek V4.1 Flash 1017 performance at half the cost.

Why it matters: The lower cost and extended context make it attractive for building complex coding and multi‑step AI agents.

More in AI

AI17h ago2 sources

Anthropic allocates 5% of its shares to a public‑benefit fund for AI‑impacted workers

The Decoder + 1 more