MiniMax M3

by Minimax

MiniMax's frontier-class model. Strong on coding and multi-step agent work, and stays responsive across long codebases and very large prompts. Try MiniMax M3 on Zo.

Context window512k tokens
Release dateMay 2026
Prompts ≤ 512k tokens
Input
$0.30 / 1M tokens
Output
$1.20 / 1M tokens
Prompts > 512k tokens
Input
$1.20 / 1M tokens
Output
$4.80 / 1M tokens

How we rate models

Quality. Three brains is top, one is basic. We anchor the rating on SWE-bench Verifiedthe industry-standard coding benchmark — how often a model solves real GitHub issues end-to-end.
Fast. Shown when the model can produce at least roughly 80 tokens per second (about 60 words a second). Not every model qualifies.
$$$$$$$$$$
Cost. Single dollar is the cheapest tier (green), four dollars is the priciest (amber) — reserved for premium models that cost well above the rest. Tiers sit on a fixed scale of what a typical request costs, so they're comparable across providers and don't shift as the lineup changes.

On Zo, you can choose between all the leading AI models for chatting, research, coding, and generating media. We believe in giving you the power to explore different models so that you can get the best results for any task.

MiniMax M3 | Zo Computer