MiniMax M3.1 Flash: a very cheap model for running agents, with one big caveat

Chinese lab MiniMax released M3.1 Flash, an everyday model for coding, research and agent work with a 1 million token context window, priced at $0.15 per million input tokens and $0.60 per million output tokens. MiniMax says it runs nearly three times faster than its previous model and cuts the cost of agent tasks by 70%; those are the company's own figures. Independent testers Artificial Analysis put it close to DeepSeek's new Flash model for less money, but also found it is unusually likely to give a wrong answer rather than admit it doesn't know.

Why it mattersIf you run high-volume jobs like sorting, drafting or coding tasks, this could cut your AI bill a lot. Use it where answers are checked (code with tests, drafts a person reviews), not for facts you'll publish or send to customers unchecked.

How we knowCompany figures are labelled as their own claims; anything we couldn't verify is labelled as a report or a claim.