MiniMax: MiniMax M3 (batch)

MiniMax M3 (batch) is a model from MiniMax that accepts text, image, and video input and returns text. It offers a 524,288-token context window and allows completions of up to 512,000 tokens, so long documents and lengthy generated outputs fit within a single request. It supports tool calling and reasoning; structured output support is not listed, so verify it before relying on schema-constrained responses. At $0.224 per million input tokens and $0.88 per million output tokens, it sits at the low end of paid pricing, which suits high-volume or long-context workloads. Its blended benchmark score of 78.0 rests on only three benchmarks, so treat the standing as provisional. Within that set, the split is uneven: 96.6 on the coding index, 73.3 on the general index, and 58.3 on the agentic index. Shortlist MiniMax M3 (batch) for cost-sensitive coding work, and test it carefully before using it for multi-step agentic tasks.