AI NexusAI Nexus
Model release

MiniMax launches M3.1 Flash for hosted agent and coding workflows

Event time 1 independent sourceEditorial score 77/100Updated here

The short version

MiniMax has released M3.1 Flash through its API, M Plan and MiniMax Code, listing flat input, output and cached-input token prices; weights are planned for a later release.

What changed

M3.1 Flash is a new hosted MiniMax model for agentic tasks, coding and computer use, with published $0.15 input, $0.60 output and $0.015 cached input pricing per million tokens.

What it means for you

Teams evaluating hosted coding agents have a new MiniMax model to benchmark, but should not plan for self-hosting until the promised weights and license are actually released.

MiniMax has introduced M3.1 Flash, a hosted model the company positions for agentic work, coding and computer use. The official launch page says it is available through the MiniMax API, M Plan and MiniMax Code.

MiniMax publishes flat API rates of $0.15 per million input tokens, $0.60 per million output tokens and $0.015 per million cached-input tokens. Those are model-token prices, not a complete cost forecast: teams should also measure retries, tool execution, context retention and any third-party service costs in their own workflow.

The company reports benchmark results and latency figures, but those are vendor claims rather than an independent comparison. Start with a bounded, read-only evaluation and compare completion, latency and recovery behavior with the model already in use.

The release is a hosted route today. MiniMax says weights will be released later, so teams that require a self-controlled deployment boundary should wait for a specific package and license rather than infer availability from the announcement.

Related catalog entry

Fact check

  • VerifiedMiniMax announced M3.1 Flash for agentic tasks, coding and computer use, and says it is available through the MiniMax API, M Plan and MiniMax Code.Evidence
  • VerifiedMiniMax lists M3.1 Flash at $0.15 input, $0.60 output and $0.015 cached input per million tokens, and says model weights will be released later.Evidence

Coverage timeline

  1. Primary sourceMiniMax
    MiniMax M3.1 Flash: Redefining Agent Performance and Cost