MiniMax M3
Release date: June 1, 2026. MiniMax positions M3 as a natively multimodal model for coding and agentic work, with a context window of up to 1 million tokens. (minimax.io)
Capabilities and uses
- Accepts text, image and video inputs. Its sparse-attention architecture, MiniMax Sparse Attention (MSA), is designed to improve long-context efficiency.
- Intended for software development, debugging, repository understanding, search, office workflows and longer-running agent tasks. Computer use is available through the MiniMax Code application.
- The model card describes approximately 428 billion total parameters and 23 billion activated parameters. It lists enabled, adaptive and disabled reasoning modes.
- Model weights are available under the MiniMax Community License; review its terms before deployment. (minimax.io)
Limitations
The context window is limited to up to 1 million tokens, and requests above 512k input tokens have a higher API rate. Agent and computer-use workflows depend on the surrounding application and tools; outputs and actions should be reviewed, particularly for consequential tasks. The published model weights are subject to their license terms. (minimax.io)
Official API pricing
The following are pay-as-you-go rates for MiniMax-M3, checked on November 16, 2026. Every figure is per 1 million tokens. Prices shown are the official listed rates after the stated permanent 50% discount. Input-length thresholds apply per request; cached-input reads have the separate rate shown. Priority service costs 1.5 times the standard rate and is enabled with service_tier=priority. The domestic and international platforms are separate offerings, not currency conversions of one another. (platform.minimax.cn)
Mainland China platform — CNY
| Service | Input length | Input | Output | Cache read |
|---|---|---|---|---|
| Standard | ≤512k tokens | ¥2.10 | ¥8.40 | ¥0.42 |
| Standard | >512k tokens | ¥4.20 | ¥16.80 | ¥0.84 |
| Priority | ≤512k tokens | ¥3.15 | ¥12.60 | ¥0.63 |
| Priority | >512k tokens | ¥6.30 | ¥25.20 | ¥1.26 |
Source: MiniMax China pricing overview, checked November 16, 2026. (platform.minimax.cn)
International platform — USD
| Service | Input length | Input | Output | Cache read |
|---|---|---|---|---|
| Standard | ≤512k tokens | $0.30 | $1.20 | $0.06 |
| Standard | >512k tokens | $0.60 | $2.40 | $0.12 |
| Priority | ≤512k tokens | $0.45 | $1.80 | $0.09 |
| Priority | >512k tokens | $0.90 | $3.60 | $0.18 |
Source: MiniMax international pricing overview, checked November 16, 2026. (platform.minimax.io)