moonshotai/kimi-k2.5
Kimi K2.5 is Moonshot AI's native multimodal model, delivering state-of-the-art visual coding capability and a self-directed agent swarm paradigm. Built on Kimi K2 with continued pretraining over approximately 15T mixed...
| Provider | In $/M | Out $/M | Context | Uptime |
|---|---|---|---|---|
| DigitalOcean | $0.375 | $2.02 | 256K | 99.7% |
| ModelRunfp4 | $0.400 | $1.90 | 262K | 98.4% |
| Chutesint4 | $0.440 | $2.00 | 262K | 84.9% |
| DeepInfrafp4 | $0.450 | $2.25 | 262K | 99.9% |
| SiliconFlowint4 | $0.450 | $2.25 | 262K | 99.8% |
| AtlasCloudint4 | $0.490 | $2.50 | 262K | 99.7% |
| StreamLake | $0.540 | $2.70 | 256K | 99.7% |
| Venice | $0.560 | $3.50 | 256K | — |
| Novita | $0.570 | $2.85 | 262K | 99.9% |
| Phala | $0.600 | $3.00 | 262K | — |
| Moonshot AIint4 | $0.600 | $3.00 | 262K | 100% |
| Fireworks | $0.600 | $3.00 | 262K | 0% |
Kimi K2.5 costs $0.375 per million input tokens and $2.02 per million output tokens via OpenRouter, making it 143rd cheapest of 298 paid models.
Kimi K2.5 scores 37.3 on the Artificial Analysis Intelligence Index, ranking 53rd of 178 benchmarked models, with a GPQA Diamond score of 79%.
Kimi K2.5 generates around 81 tokens per second with 218ms time-to-first-token (p50), the 91st fastest tracked model.
Kimi K2.5 supports a 262K-token context window. It accepts text, image input.