Kimi K3 is Kimi’s most capable flagship model to date, with 2.8 trillion parameters. It is built on Kimi Delta Attention (KDA), a hybrid linear attention mechanism, and Attention Residuals, with native visual understanding and a 1M-token context window. It is the world’s first open-source model in the 3-trillion-parameter class, designed for frontier intelligence scenarios including long-horizon coding, knowledge work, and reasoning.
同系列或更低输入价的公开模型。 / Same provider or cheaper catalog alternatives.
| Model | Provider | Type | Input | Output |
|---|---|---|---|---|
| MiniMax-M2.1 | MiniMax | chat | 2.73 Token / 1M tokens | 10.92 Token / 1M tokens |
| MiniMax-M2.1-highspeed | MiniMax | chat | 5.46 Token / 1M tokens | 21.84 Token / 1M tokens |
| MiniMax-M2.5 |
A quick read of the model identity, commercial setup, and access footprint.
Published rates, tier transitions, and modality-specific surcharges in the current catalog entry.
| Pricing | Billing mode |
|---|---|
| Prompt | $3.0000/1M tokens |
| Output | $15.0000/1M tokens |
| Cache hit | $0.3000/1M tokens |
Hourly usage, latency, reliability, and throughput for this model.
Top 10 AI coding agents using this model during the last 15 days.
Top 10 official SDKs and language HTTP clients using this model during the last 15 days.
| MiniMax |
| chat |
| 0.2730 Token / 1M tokens |
| 10.92 Token / 1M tokens |
| MiniMax-M2.5-highspeed | MiniMax | chat | 5.46 Token / 1M tokens | 21.84 Token / 1M tokens |
| MiniMax-M2.7 | MiniMax | chat | 2.73 Token / 1M tokens | 10.92 Token / 1M tokens |
| MiniMax-M2.7-highspeed | MiniMax | chat | 5.46 Token / 1M tokens | 16.80 Token / 1M tokens |
