Kimi K3
Multimodal Kimi model with 1M context and toggleable max-effort thinking for long-horizon agent work
Shogo Agent Index
Supported#26 of 9254.9range 47.3–62.5 · 29% of weight measured
Agentic coding
63
31% measured
Long-horizon autonomy
65
50% measured
Knowledge-work agents
42
32% measured
Computer use
43
estimated
Security & context learning
53
estimated
Agentic benchmarks
| Benchmark | Score | Normalised | Index weight | Harness / source |
|---|---|---|---|---|
| SWE-bench Verified | est. | 8% | predicted 82 | |
| Terminal-Bench 2.0 | est. | 8% | predicted 83 | |
| DeepSWE | 68.5% | 6% | mini-swe-agent · $4.65/task | |
| FrontierSWE | 25.9% | 5% | proximus · $109.71/task | |
| CursorBench | est. | 4% | predicted 42 | |
| GSO-Bench | est. | 4% | predicted 42 | |
| METR Time Horizons | est. | 10% | predicted 79 | |
| Vending-Bench 2 | $5,165 | 6% | source | |
| PostTrainBench | 31.5% | 4% | Claude Code | |
| APEX-Agents | 47.8% | 8% | source | |
| GDPval | est. | 7% | predicted 54 | |
| Remote Labor Index | est. | 5% | predicted 6 | |
| DeepResearch Bench | est. | 5% | predicted 49 | |
| OSWorld 2.0 | est. | 6% | predicted 16 | |
| OSWorld | est. | 4% | predicted 84 | |
| Cybench | est. | 5% | predicted 82 | |
| CL-bench | est. | 5% | predicted 24 |
Other benchmarks
| Benchmark | Category | Score | Source |
|---|---|---|---|
| ARC-AGI | Reasoning | 94.5% | source |
| ARC-AGI-2 | Reasoning | 60.4% | source |
| Chess Puzzles | Reasoning | 37.0% | source |
| CritPt | Reasoning | 23.4% | source |
| DTBench | Knowledge | 91.2% | source |
| FrontierCode | Coding | 44.2% | mini-swe-agent |
| FrontierMath Tiers 1-3 | Math | 72.2% | source |
| FrontierMath Tier 4 | Math | 39.0% | source |
| GPQA Diamond | Reasoning | 93.1% | source |
| LMCA | Knowledge | 52.7% | source |
| MirrorCode | Coding | 54.4% | source |
| OTIS Mock AIME | Math | 97.2% | source |
| ProofBench | Math | 87.0% | source |
| SciCode | Coding | 59.5% | source |
| SimpleBench | Reasoning | 60.7% | source |
| SimpleQA Verified | Knowledge | 50.6% | source |
| WebDev Arena | Coding | 1674 | source |
| WeirdML v3 | Coding | 7.5% | kimi_code |
API hosts
Live prices and uptime for every host serving Kimi K3 via OpenRouter.
| Host | Input | Output | Cache read | Agent $/M | Uptime 24h | Quant |
|---|---|---|---|---|---|---|
| Wafer | $0.28 | $9.60 | $0.13 | $0.60 | 100.0% | mxfp4 |
| Relace | $0.18 | $11.01 | $0.18 | $0.70 | 99.9% | fp4 |
| Chutes | $1.85 | $9.50 | $0.18 | $0.87 | 96.9% | mxfp4 |
| Wafer | $0.15 | $16.00 | $0.15 | $0.91 | 100.0% | mxfp4 |
| InferenceNet | $0.24 | $15.00 | $0.21 | $0.92 | 100.0% | fp4 |
| InferenceNet | $0.24 | $15.00 | $0.21 | $0.92 | 100.0% | fp4 |
| Wafer | $0.19 | $16.00 | $0.18 | $0.94 | 100.0% | mxfp4 |
| Sail Research | $0.34 | $16.58 | $0.17 | $0.98 | 100.0% | fp4 |
| KT | $0.45 | $15.00 | $0.30 | $1.02 | 99.1% | — |
| Io Net | $1.90 | $14.00 | $0.19 | $1.09 | 99.3% | fp4 |
| DigitalOcean | $2.55 | $12.95 | $0.26 | $1.19 | 100.0% | — |
| InferenceNet | $0.45 | $18.00 | $0.35 | $1.20 | 100.0% | fp4 |
| InferenceNet | $0.45 | $18.00 | $0.35 | $1.20 | 100.0% | fp4 |
| DeepInfra | $2.85 | $14.25 | $0.28 | $1.32 | 99.8% | mxfp4 |
| Alibaba | $3.00 | $15.00 | $0.30 | $1.39 | 100.0% | — |
| Amazon Bedrock | $3.00 | $15.00 | $0.30 | $1.39 | 100.0% | — |
| BaseTen | $3.00 | $15.00 | $0.30 | $1.39 | 99.7% | fp8 |
| Fireworks | $3.00 | $15.00 | $0.30 | $1.39 | 99.8% | — |
| Modal | $3.00 | $15.00 | $0.30 | $1.39 | 100.0% | mxfp4 |
| Moonshot AI | $3.00 | $15.00 | $0.30 | $1.39 | 99.4% | — |
| Nebius | $3.00 | $15.00 | $0.30 | $1.39 | 100.0% | fp4 |
| Parasail | $3.00 | $15.00 | $0.30 | $1.39 | 100.0% | fp4 |
| River | $3.00 | $15.00 | $0.30 | $1.39 | 99.9% | — |
| Together | $3.00 | $15.00 | $0.30 | $1.39 | 98.4% | — |
| Makora | $1.00 | $18.00 | $0.50 | $1.40 | 100.0% | fp4 |
| Amazon Bedrock | $3.30 | $16.50 | $0.33 | $1.52 | 100.0% | — |
| Fireworks | $4.50 | $22.50 | $0.45 | $2.08 | 96.7% | — |
| Fireworks | $4.50 | $22.50 | $0.45 | $2.08 | 98.9% | — |
| Parasail | $4.50 | $22.50 | $0.45 | $2.08 | 100.0% | fp4 |