Gemini 3.6 Flash vs Kimi K3: Benchmark Comparison
Gemini 3.6 Flash has the higher score on 10 of 24 shared benchmarks; Kimi K3 leads on 14.
The largest observed score gap is 29.88 pts on CyberBench v1.1 , where Kimi K3 leads.
Reported ±1 standard-error ranges overlap on 11 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 | Gemini 3.6 Flash | Kimi K3 | Gap | Reported uncertainty |
|---|---|---|---|---|
| Harvey's Legal Agent Benchmark | 3.33% ±1.17 | 12.92% ±2.68 | 9.58 pts | Reported ±1 SE ranges do not overlap |
| Legal Research Bench | 25.00% ±3.01 | 46.15% ±3.46 | 21.15 pts | Reported ±1 SE ranges do not overlap |
| LegalBench | 86.70% ±0.41 | 86.21% ±0.41 | 0.49 pts | Reported ±1 SE ranges overlap |
| EMB | 65.41% ±2.65 | 66.68% ±2.87 | 1.27 pts | Reported ±1 SE ranges overlap |
| Finance Agent (v2) | 56.30% ±0.18 | 53.11% ±0.37 | 3.18 pts | Reported ±1 SE ranges do not overlap |
| MortgageTax | 67.61% ±0.92 | 66.34% ±0.91 | 1.27 pts | Reported ±1 SE ranges overlap |
| Tax Agent Bench | 49.90% ±3.30 | 68.67% ±3.08 | 18.78 pts | Reported ±1 SE ranges do not overlap |
| TaxEval v2 | 74.86% ±0.85 | 75.72% ±0.84 | 0.86 pts | Reported ±1 SE ranges overlap |
| MedCode | 53.15% ±2.16 | 49.36% ±2.20 | 3.80 pts | Reported ±1 SE ranges overlap |
| MedScribe | 79.66% ±1.86 | 88.05% ±1.98 | 8.39 pts | Reported ±1 SE ranges do not overlap |
| BioMysteryBench | 58.52% ±1.61 | 72.59% ±1.61 | 14.07 pts | Reported ±1 SE ranges do not overlap |
| Terminal-Bench Science | 4.29% ±2.44 | 1.43% ±1.43 | 2.86 pts | Reported ±1 SE ranges overlap |
| GPQA Diamond | 93.43% ±1.33 | 92.93% ±1.31 | 0.50 pts | Reported ±1 SE ranges overlap |
| MMLU Pro | 89.28% ±0.30 | 87.97% ±0.32 | 1.30 pts | Reported ±1 SE ranges do not overlap |
| MMMU Pro | 88.38% ±0.77 | 88.15% ±0.78 | 0.23 pts | Reported ±1 SE ranges overlap |
| SAGE | 48.69% ±3.38 | 52.78% ±3.41 | 4.10 pts | Reported ±1 SE ranges overlap |
| Code Migration | 30.93% ±4.07 | 16.10% ±4.10 | 14.83 pts | Reported ±1 SE ranges do not overlap |
| IOI | 35.06% ±5.70 | 48.94% ±9.82 | 13.89 pts | Reported ±1 SE ranges overlap |
| LiveCodeBench | 88.08% ±0.94 | 87.19% ±0.97 | 0.89 pts | Reported ±1 SE ranges overlap |
| ProgramBench | 0.00% ±0.00 | 2.00% ±0.99 | 2.00 pts | Reported ±1 SE ranges do not overlap |
| SWE-bench | 79.60% ±1.80 | 93.40% ±1.11 | 13.80 pts | Reported ±1 SE ranges do not overlap |
| Vibe Code Bench v1.1 | 64.00% ±4.29 | 84.97% ±2.75 | 20.96 pts | Reported ±1 SE ranges do not overlap |
| CyberBench v1.1 | 45.36% ±3.75 | 75.24% ±5.56 | 29.88 pts | Reported ±1 SE ranges do not overlap |
| Public Benefits Bench v1.1 | 56.56% ±1.29 | 68.20% ±1.21 | 11.64 pts | Reported ±1 SE ranges do not overlap |
Performance by category
| Category | Gemini 3.6 Flash average | Kimi K3 average |
|---|---|---|
| Legal | 38.35% | 48.43% |
| Finance | 62.81% | 66.10% |
| Healthcare | 66.41% | 68.70% |
| Science | 31.40% | 37.01% |
| Academic | 90.36% | 89.68% |
| Education | 48.69% | 52.78% |
| Coding | 49.61% | 55.43% |
| Cyber | 45.36% | 75.24% |
| Social Mobility | 56.56% | 68.20% |
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 | Gemini 3.6 Flash cost | Kimi K3 cost | Gemini 3.6 Flash latency | Kimi K3 latency |
|---|---|---|---|---|
| Harvey's Legal Agent Benchmark | $1.90 | $3.83 | 16m24s | 13m58s |
| Legal Research Bench | $0.47 | $3.47 | 2m48s | 12m40s |
| LegalBench | N/A | N/A | 2.96s | 7.11s |
| EMB | $3.89 | $3.07 | 12m06s | 12m21s |
| Finance Agent (v2) | $1.40 | $1.91 | 3m45s | 4m46s |
| MortgageTax | N/A | N/A | 9.24s | 78.94s |
| Tax Agent Bench | $0.36 | $2.79 | 105.32s | 40m15s |
| TaxEval v2 | N/A | N/A | 12.34s | 2m25s |
| MedCode | N/A | N/A | 16.68s | 36.99s |
| MedScribe | N/A | N/A | 33.55s | 38.24s |
| BioMysteryBench | $1.25 | $1.06 | 12m42s | 4m55s |
| Terminal-Bench Science | $4.39 | $21.30 | 1h52m | 4h49m |
| GPQA Diamond | N/A | N/A | 18.06s | 2m02s |
| MMLU Pro | N/A | N/A | 7.30s | 38.72s |
| MMMU Pro | N/A | N/A | 11.31s | 106.52s |
| SAGE | N/A | N/A | 21.96s | 47.30s |
| Code Migration | $8.93 | $13.87 | 1h30m | 4h18m |
| IOI | $8.67 | $14.17 | 48m23s | 4h18m |
| LiveCodeBench | N/A | N/A | 31.88s | 3m20s |
| ProgramBench | $5.92 | $70.48 | 47m33s | 5h42m |
| SWE-bench | $1.19 | $0.76 | 13m08s | 10m19s |
| Vibe Code Bench v1.1 | $3.04 | $10.01 | 25m23s | 16m39s |
| CyberBench v1.1 | $1.31 | $2.13 | 11m00s | 30m44s |
| Public Benefits Bench v1.1 | $0.39 | $0.91 | 2m02s | 7m19s |
Results available only for Gemini 3.6 Flash
None.
Results available only for Kimi K3
- Vals Index
- Vals RSI Index
- ProofBench v1.1
- Terminal-Bench 4.0
- Vibe Code Bench 1-100
- Time Horizon Index: KSP