Kimi K3 vs GPT-6 Luna: Benchmark Comparison
Kimi K3 has the higher score on 20 of 24 shared benchmarks; GPT-6 Luna leads on 4.
The largest observed score gap is 23.00 pts on ProofBench v1.1 , where Kimi K3 leads.
Reported ±1 standard-error ranges overlap on 8 of 24 shared benchmarks with comparable uncertainty data. This is not a pairwise statistical significance test.
Shared benchmark results
Scores and reported standard errors come from the latest visible Vals benchmark version available for both models.
| Benchmark | Kimi K3 | GPT-6 Luna | Gap | Reported uncertainty |
|---|---|---|---|---|
| Vals Index | 52.95% ±1.00 | 52.32% ±1.12 | 0.63 pts | Reported ±1 SE ranges overlap |
| Vals Finance Index | 57.95% ±1.43 | 53.05% ±1.43 | 4.90 pts | Reported ±1 SE ranges do not overlap |
| Vals Legal Index | 71.72% ±1.80 | 62.20% ±1.57 | 9.52 pts | Reported ±1 SE ranges do not overlap |
| HLAB-Vals-1.1 | 12.92% ±2.81 | 4.17% ±1.63 | 8.75 pts | Reported ±1 SE ranges do not overlap |
| Immigration | 68.26% ±1.51 | 62.53% ±1.57 | 5.73 pts | Reported ±1 SE ranges do not overlap |
| Legal Research Bench | 46.15% ±3.46 | 30.29% ±3.19 | 15.87 pts | Reported ±1 SE ranges do not overlap |
| EMB v1.1 | 71.26% ±2.28 | 69.95% ±2.54 | 1.31 pts | Reported ±1 SE ranges overlap |
| Finance Agent v2.1 | 60.01% ±2.09 | 56.84% ±2.12 | 3.17 pts | Reported ±1 SE ranges overlap |
| Tax Agent Bench | 33.37% ±3.39 | 19.95% ±2.87 | 13.42 pts | Reported ±1 SE ranges do not overlap |
| MedCode | 49.36% ±2.20 | 44.69% ±2.30 | 4.67 pts | Reported ±1 SE ranges do not overlap |
| MedScribe | 88.05% ±1.98 | 83.71% ±1.95 | 4.34 pts | Reported ±1 SE ranges do not overlap |
| ProofBench v1.1 | 87.00% ±3.38 | 64.00% ±4.82 | 23.00 pts | Reported ±1 SE ranges do not overlap |
| BioMysteryBench | 72.59% ±1.61 | 61.48% ±2.59 | 11.11 pts | Reported ±1 SE ranges do not overlap |
| Terminal-Bench-Science | 1.43% ±1.43 | 4.29% ±2.44 | 2.86 pts | Reported ±1 SE ranges overlap |
| SAGE | 86.77% ±1.85 | 83.40% ±2.30 | 3.37 pts | Reported ±1 SE ranges overlap |
| Code Migration v1.1 | 28.59% ±3.78 | 46.67% ±3.61 | 18.08 pts | Reported ±1 SE ranges do not overlap |
| IOI | 48.94% ±9.82 | 55.56% ±8.87 | 6.61 pts | Reported ±1 SE ranges overlap |
| ProgramBench | 2.00% ±0.99 | 0.50% ±0.50 | 1.50 pts | Reported ±1 SE ranges do not overlap |
| SWE-bench Multimodal | 44.17% ±2.27 | 37.50% ±2.21 | 6.67 pts | Reported ±1 SE ranges do not overlap |
| Terminal-Bench 4.0 | 17.17% ±0.51 | 13.64% ±1.51 | 3.54 pts | Reported ±1 SE ranges do not overlap |
| Vibe Code Bench v1.1 | 84.97% ±2.75 | 81.65% ±3.38 | 3.32 pts | Reported ±1 SE ranges overlap |
| CyberBench v1.1 | 75.24% ±5.56 | 76.25% ±5.29 | 1.01 pts | Reported ±1 SE ranges overlap |
| Financial Coaching Bench | 74.53% ±1.49 | 65.24% ±1.68 | 9.28 pts | Reported ±1 SE ranges do not overlap |
| Public Benefits Bench v1.1 | 68.20% ±1.21 | 57.65% ±1.28 | 10.55 pts | Reported ±1 SE ranges do not overlap |
Performance by category
| Category | Kimi K3 average | GPT-6 Luna average |
|---|---|---|
| Legal | 42.44% | 32.33% |
| Finance | 54.88% | 48.91% |
| Healthcare | 68.70% | 64.20% |
| Math | 87.00% | 64.00% |
| Science | 37.01% | 32.88% |
| Education | 86.77% | 83.40% |
| Coding | 37.64% | 39.25% |
| Cyber | 75.24% | 76.25% |
| Social Mobility | 71.36% | 61.44% |
Cost and latency
Cost per test appears only where the benchmark reports it for both models. Latency is the measured completion time for that benchmark.
| Benchmark | Kimi K3 cost | GPT-6 Luna cost | Kimi K3 latency | GPT-6 Luna latency |
|---|---|---|---|---|
| Vals Index | $7.17 | $0.45 | 1h15m | 32m23s |
| Vals Finance Index | $2.58 | $0.15 | 22m24s | 22m01s |
| Vals Legal Index | $3.47 | $0.31 | 31m51s | 32m30s |
| HLAB-Vals-1.1 | $4.14 | $0.30 | 42m36s | 21m45s |
| Immigration | $0.18 | $0.01 | 64.84s | 2m31s |
| Legal Research Bench | $3.47 | $0.44 | 12m40s | 42m28s |
| EMB v1.1 | $3.07 | $0.14 | 12m21s | 17m21s |
| Finance Agent v2.1 | $1.88 | $0.12 | 14m35s | 15m27s |
| Tax Agent Bench | $2.79 | $0.19 | 40m15s | 33m16s |
| MedCode | N/A | N/A | 36.99s | 108.77s |
| MedScribe | N/A | N/A | 38.24s | 2m50s |
| ProofBench v1.1 | $1.65 | $0.04 | 34m31s | 7m31s |
| BioMysteryBench | $1.06 | $0.05 | 4m55s | 8m44s |
| Terminal-Bench-Science | $21.30 | $0.24 | 4h49m | 1h16m |
| SAGE | $0.23 | $0.01 | 2m25s | 47.63s |
| Code Migration v1.1 | $35.75 | $0.83 | 4h31m | 1h00m |
| IOI | $14.17 | $0.20 | 4h18m | 35m47s |
| ProgramBench | $70.48 | $0.18 | 5h42m | 28m55s |
| SWE-bench Multimodal | $1.65 | $0.03 | 22m45s | 5m16s |
| Terminal-Bench 4.0 | $12.48 | $0.35 | 3h09m | 32m46s |
| Vibe Code Bench v1.1 | $10.01 | $1.35 | 16m39s | 37m00s |
| CyberBench v1.1 | $2.13 | $0.13 | 30m44s | 15m50s |
| Financial Coaching Bench | $0.19 | $0.02 | 8m29s | 5m49s |
| Public Benefits Bench v1.1 | $0.91 | $0.22 | 7m19s | 52m59s |
Results available only for Kimi K3
- Vals RSI Index
- Vals Inference Provider Index
- LegalBench
- MortgageTax
- TaxEval v2
- GPQA Diamond
- MMLU Pro
- MMMU Pro
- LiveCodeBench
- SWE-bench
- CyberOffense
- Access to Justice Benchmark
- Time Horizon Index: KSP
Results available only for GPT-6 Luna
- Vals Science Index
- ClaimBench
- MysteryMechanism
- SRE Bench