MiniMax M3

MiniMax's open-weight flagship (June 1, 2026): a ~428B-parameter MoE (23B active) using MiniMax Sparse Attention for efficient 1M-token context. Strong coding, agentic and computer-use performance at roughly 5–10% of closed-frontier cost; a ~50% launch promo applied at release. Benchmarks are vendor-reported — verify.

MiniMax M3 strengths

• Open weights (self-hostable)
• 1M context via sparse attention
• Frontier coding + agentic / computer-use
• Native multimodal (incl. video input)
• Very low cost vs closed frontier models

Pricing & context

Cost guide: a typical call of about 10K input + 2K output tokens costs roughly $0.011 at list prices. Worth modelling against cheaper tiers before committing high-volume traffic.

When to choose MiniMax M3

MiniMax M3 is best for long-context coding agents, computer-use automation and high-volume tasks wanting frontier quality with open weights. If your workload is more cost-sensitive, weigh it against gpt-oss-120b (≈$0.03 input /1M) first.

MiniMax M3 FAQ

How much does MiniMax M3 cost?

MiniMax M3 is priced at $0.60 per 1M input tokens and $2.40 per 1M output tokens (public API list price), with a 1M tokens (min 512K) context window. A typical call of about 10K input and 2K output tokens costs roughly $0.011.

What is MiniMax M3 best for?

MiniMax M3 by MiniMax is best for long-context coding agents, computer-use automation and high-volume tasks wanting frontier quality with open weights.

How does MiniMax M3 pricing compare to Doubao Seed 2.1 Pro?

MiniMax M3 input costs $0.60 per 1M tokens versus $0.83 for Doubao Seed 2.1 Pro, roughly 1.4x less expensive on input. Output is $2.40 vs $4.17.

Is MiniMax M3 multimodal?

MiniMax M3 supports text, image, video.