Prices use the model developer's current official list prices. Billing units and applicable conditions are preserved for each tier; provider quotes may differ.
| Field | ||
|---|---|---|
| Organization | Moonshot AI | |
| Family | Kimi K3 | |
| Released | - | |
| Context window | 1M | |
| Max output | 1M | |
| Input modality | TextImageVideo | |
| Output type | Text | |
| Providers | ai&, Atlas Cloud, Baseten, Charm Hyper, Fireworks AI, KIE AI, Liquid AI, Moonshot AI (China), Moonshot AI (Global), Morph, Neuralwatt, OpenCode Go, OpenRouter, Tinfoil, Wafer | |
| Input | $2 / 1M tokens | |
| Output | $10 / 1M tokens | |
| Cache Read | $0.2 / 1M tokens | |
| Cache Write | $2 / 1M tokens | |
| Other charges | - | |
| Weights | Open | |
| Reasoning | ||
| Structured | ||
| Tool call | ||
| Open weight |
Performance
Scoring rulesText scores use the site’s full-population calibration and scoring method, on a 0–100 scale. Media results stay separate by task and configuration; they do not contribute to the text score.