Real human-preference Elo from LMArena blind head-to-head votes. Higher is better; — means not yet ranked in that arena. This is measured, not our estimate.
Blended $/1M across tracked versions of this line.
Typical 3:1 output-to-input mix, per 1M tokens
Price as of 2026-05-10 · Source: minimax_official_pricing
Mixed text, image, audio and document workloads that benefit from one model across modalities.
MiniMax M3 is MiniMax's frontier multimodal coding and agent model with a 1M-token context window.
minimax-m3 is a Multimodal model from MiniMax (CN). HotON.ai tracks it at $0.30 per 1M input tokens and $1.20 per 1M output tokens, with a 1000K-token context window. Its composite efficiency score is 96/100 at an estimated $0.001 per successful task.
minimax-m3 is tracked at $0.30 per 1M input tokens and $1.20 per 1M output tokens. A typical 3:1 output-to-input workload blends to roughly $0.97 per 1M tokens. Figures are illustrative demo data.
Mixed text, image, audio and document workloads that benefit from one model across modalities.
minimax-m3 supports up to a 1000K-token context window — large enough for long documents and extended conversations in a single request.
Within the HotON.ai tracked set, minimax-m3 is cheaper than 50% of models on input price and ranks #24 of 521 by overall efficiency.
Yes — gpt-4.1-nano is a lower-cost option at $0.40 per 1M output tokens, while still covering similar Multimodal use cases. Compare them side by side on HotON.ai.
Pricing is real (via the TestKey catalog, updated daily). Quality (Arena Elo) is real where the model is ranked on LMArena. Speed, availability and efficiency are modeled estimates.