Model Release Radar
Price/capability frontier position, benchmark scores, and community signal
Model Release Radar
Kimi K3
moonshot·open weights·Kimi K3 license·released Thursday, Jul 16, 2026
Token prices
USD per million tokens via OpenRouter, Moonshot AI (moonshotai/mxfp4). Checked 2026-10-17 08:54 UTC.
| Token type | USD /1M |
|---|---|
| Input | $3 |
| Output | $15 |
| Cache read | $0.3 |
| Cache write (default) | Unknown |
| Cache write (1 hour) | Unknown |
Cache-write rates with a verified basis replace the ordinary input charge. Unknown rates are never treated as zero.
Frontier position
- Behind frontier AA coding index (blended per-1M-token price, 3:1 input/output; not a measured per-task cost) - Qwen3.8 Max (0902), Grok 4.6, Muse Spark 1.3, Gemini 3.8 Flash are priced the same or lower and at least as capable.
- Behind frontier AA intelligence index (blended per-1M-token price, 3:1 input/output; not a measured per-task cost) - Gemini 4 Argon, Claude Sonnet 5.5, GPT-6.1 Sol, GPT-6 Sol, Qwen3.8 Max (0902) are priced the same or lower and at least as capable.
- Behind frontier DeepSWE pass@1 (agentic coding) (measured USD per task) - GLM-5.3, GPT-5.6 Sol, Claude Opus 5, Gemini 3.8 Flash are priced the same or lower and at least as capable.
Cost vs. capability
Loading chart…
Benchmark scores
Measured cost per task (DeepSWE)
| Benchmark | Pass@1 | Measured cost / task | Frontier |
|---|---|---|---|
| DeepSWE (agentic coding) (max effort, 95% CI 64.0%-73.0%, 4 runs) | 68.5% | $4.65 (median $3.60) | behind |
Artificial Analysis scores
Artificial Analysis scores describe capability. Compare them against estimated token spend with your cache mix; these estimates are not measured task costs.
| Benchmark | Score |
|---|---|
| AA intelligence index | 43.6 |
| AA coding index | 76.2 |
Community signal
| Variant | Coding Elo | Overall Elo | Votes |
|---|---|---|---|
| Rating | 1,507 | 1,475 | 32,795 |