MiniMax M3.1 Flash

As of October 26, 2026, MiniMax M3.1 Flash by MiniMax costs $0.15 per 1M input tokens and $0.60 per 1M output tokens, with a 1M-token context window.

MiniMax M3.1 Flash went generally available on October 19, 2026, after a preview inside MiniMax Code on September 27, and the pitch is price: $0.15 per million input tokens, $0.60 output, and $0.015 on a cache hit, which is half the standard M3 rate on every line. MiniMax positions it for agentic work, coding, computer use, and cybersecurity, with chained tool calls as the shape it is tuned for. It takes text, image, and video input across a 1 million token context window and always reasons before answering, with the reasoning depth adjustable rather than switchable off. MiniMax reports 73.7 percent on DeepSWE v1.1 at under $0.30 per task; that figure is vendor-reported. The independent cost reading is more useful: Artificial Analysis puts its Intelligence Index task cost at $0.22 against $0.51 for M3, a 57 percent drop, which lines up with the halved list price rather than exceeding it. Weights are not out. MiniMax says they will be released later, so unlike M3 this is a hosted-only model for now, and the earlier preview id is distinct from the general-availability one, which is worth checking if you wired up the preview in September.