DeepSeek V4.1 Flash vs Kimi K3: Benchmark Comparison
DeepSeek V4.1 Flash has the higher score on 4 of 21 shared benchmarks; Kimi K3 leads on 17.
The largest observed score gap is 33.00 pts on ProofBench v1.1 , where Kimi K3 leads.
Reported ±1 standard-error ranges overlap on 11 of 21 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 | DeepSeek V4.1 Flash | Kimi K3 | Gap | Reported uncertainty |
|---|---|---|---|---|
| Vals Index | 51.32% ±1.13 | 50.30% ±0.99 | 1.02 pts | Reported ±1 SE ranges overlap |
| Harvey's Legal Agent Benchmark | 6.67% ±1.83 | 12.92% ±2.68 | 6.25 pts | Reported ±1 SE ranges do not overlap |
| Legal Research Bench | 41.35% ±3.42 | 46.15% ±3.46 | 4.81 pts | Reported ±1 SE ranges overlap |
| LegalBench | 83.28% ±0.46 | 86.21% ±0.41 | 2.94 pts | Reported ±1 SE ranges do not overlap |
| EMB | 57.21% ±3.23 | 66.68% ±2.87 | 9.47 pts | Reported ±1 SE ranges do not overlap |
| Finance Agent (v2) | 53.48% ±0.39 | 53.11% ±0.37 | 0.37 pts | Reported ±1 SE ranges overlap |
| Tax Agent Bench | 62.46% ±3.15 | 68.67% ±3.08 | 6.21 pts | Reported ±1 SE ranges overlap |
| MedCode | 41.17% ±2.04 | 49.36% ±2.20 | 8.18 pts | Reported ±1 SE ranges do not overlap |
| MedScribe | 85.50% ±1.92 | 88.05% ±1.98 | 2.55 pts | Reported ±1 SE ranges overlap |
| ProofBench v1.1 | 54.00% ±5.01 | 87.00% ±3.38 | 33.00 pts | Reported ±1 SE ranges do not overlap |
| BioMysteryBench | 67.78% ±1.11 | 72.59% ±1.61 | 4.81 pts | Reported ±1 SE ranges do not overlap |
| Terminal-Bench Science | 0.00% ±0.00 | 1.43% ±1.43 | 1.43 pts | Reported ±1 SE ranges overlap |
| SAGE | 47.88% ±3.44 | 52.78% ±3.41 | 4.90 pts | Reported ±1 SE ranges overlap |
| Code Migration | 45.62% ±4.29 | 16.10% ±4.10 | 29.52 pts | Reported ±1 SE ranges do not overlap |
| IOI | 40.28% ±2.56 | 48.94% ±9.82 | 8.67 pts | Reported ±1 SE ranges overlap |
| ProgramBench | 0.50% ±0.50 | 2.00% ±0.99 | 1.50 pts | Reported ±1 SE ranges do not overlap |
| Terminal-Bench 4.0 | 19.70% ±1.75 | 17.17% ±0.51 | 2.52 pts | Reported ±1 SE ranges do not overlap |
| Vibe Code Bench 1-100 | 16.38% ±3.19 | 18.24% ±3.92 | 1.87 pts | Reported ±1 SE ranges overlap |
| Vibe Code Bench v1.1 | 84.74% ±2.89 | 84.97% ±2.75 | 0.23 pts | Reported ±1 SE ranges overlap |
| CyberBench v1.1 | 73.69% ±5.48 | 75.24% ±5.56 | 1.55 pts | Reported ±1 SE ranges overlap |
| Public Benefits Bench v1.1 | 64.28% ±1.25 | 68.20% ±1.21 | 3.92 pts | Reported ±1 SE ranges do not overlap |
Performance by category
| Category | DeepSeek V4.1 Flash average | Kimi K3 average |
|---|---|---|
| Legal | 43.76% | 48.43% |
| Finance | 57.72% | 62.82% |
| Healthcare | 63.34% | 68.70% |
| Math | 54.00% | 87.00% |
| Science | 33.89% | 37.01% |
| Education | 47.88% | 52.78% |
| Coding | 34.54% | 31.24% |
| Cyber | 73.69% | 75.24% |
| Social Mobility | 64.28% | 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 | DeepSeek V4.1 Flash cost | Kimi K3 cost | DeepSeek V4.1 Flash latency | Kimi K3 latency |
|---|---|---|---|---|
| Vals Index | $0.33 | $6.38 | 27m53s | 1h08m |
| Harvey's Legal Agent Benchmark | $0.17 | $3.83 | 9m39s | 13m58s |
| Legal Research Bench | $0.25 | $3.47 | 11m42s | 12m40s |
| LegalBench | N/A | N/A | 6.53s | 7.11s |
| EMB | $0.23 | $3.07 | 13m00s | 12m21s |
| Finance Agent (v2) | $0.21 | $1.91 | 6m04s | 4m46s |
| Tax Agent Bench | $0.13 | $2.79 | 5m49s | 40m15s |
| MedCode | N/A | N/A | 35.13s | 36.99s |
| MedScribe | N/A | N/A | 53.83s | 38.24s |
| ProofBench v1.1 | $0.13 | $1.65 | 9m07s | 34m31s |
| BioMysteryBench | $0.23 | $1.06 | 14m50s | 4m55s |
| Terminal-Bench Science | $0.47 | $21.30 | 2h57m | 4h49m |
| SAGE | N/A | N/A | 35.78s | 47.30s |
| Code Migration | $0.94 | $13.87 | 1h45m | 4h18m |
| IOI | $0.24 | $14.17 | 18m30s | 4h18m |
| ProgramBench | $0.90 | $70.48 | 1h02m | 5h42m |
| Terminal-Bench 4.0 | $0.50 | $12.02 | 48m22s | 3h09m |
| Vibe Code Bench 1-100 | $0.69 | $9.40 | 28m21s | 1h48m |
| Vibe Code Bench v1.1 | $0.41 | $10.01 | 15m28s | 16m39s |
| CyberBench v1.1 | $0.07 | $2.13 | 13m50s | 30m44s |
| Public Benefits Bench v1.1 | $0.07 | $0.91 | 9m59s | 7m19s |
Results available only for DeepSeek V4.1 Flash
- MysteryMechanism
- SkillsBench
- SRE Bench
Results available only for Kimi K3
- Vals RSI Index
- MortgageTax
- TaxEval v2
- GPQA Diamond
- MMLU Pro
- MMMU Pro
- LiveCodeBench
- SWE-bench
- Time Horizon Index: KSP