Kimi K2.5
Moonshot AI released 2026-01-27
MiniMax M3 is both better and cheaper.
It scores +9.4 higher and costs 42% less ($0.525/M against $0.900/M) on this workload — and it does everything this model does.
DeepSeek V4 Flash 0731 is cheaper still (88% less) but drops no image input.
Every price dimension
| Input | $0.450/M |
|---|---|
| Output | $2.25/M |
| Cached input | $0.070/M |
Independent scores
| Intelligence | 36 |
|---|---|
| Coding | 46.8 |
| Agentic | 21.7 |
| LMArena Elo | 1445.2thinking |
Sources are listed separately rather than averaged. Across the 62 models both have scored they correlate at r = 0.806 — close agreement overall, but four models rank very differently between them, and a blended score would hide exactly those.
Best rankings by task
- godotgamedev #9of 43 1219
- asciiart #21of 83 1198
- mobileapps #23of 56 1198
- webapps #24of 56 1173
- fullstack #25of 59 1146
- androidnative #26of 54 1146
- website #29of 151 1269
- codecategories #32of 145 1262
Capability
| Context window | 262K |
|---|---|
| Max output | 262K |
| Input modes | text, image |
| Tool use | yes |
| Reasoning | optional |
Provenance
| Price source | openrouter.ai |
|---|---|
| Fetched | 2026-08-24 |
| Quality data | verified |
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...