DeepSeek V4 vs Kimi K2.6: Benchmark Comparison
DeepSeek V4 has the higher score on 9 of 17 shared benchmarks; Kimi K2.6 leads on 7.
The largest observed score gap is 13.36 pts on Tax Agent Bench , where DeepSeek V4 leads.
Reported ±1 standard-error ranges overlap on 10 of 17 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 | Kimi K2.6 | Gap | Reported uncertainty |
|---|---|---|---|---|
| Harvey's Legal Agent Benchmark | 3.75% ±1.43 | 1.67% ±0.83 | 2.08 pts | Reported ±1 SE ranges overlap |
| Legal Research Bench | 23.08% ±2.93 | 15.87% ±2.54 | 7.21 pts | Reported ±1 SE ranges do not overlap |
| LegalBench | 80.32% ±0.47 | 84.74% ±0.45 | 4.42 pts | Reported ±1 SE ranges do not overlap |
| EMB | 51.62% ±3.05 | 57.85% ±2.74 | 6.23 pts | Reported ±1 SE ranges do not overlap |
| Finance Agent (v2) | 44.08% ±0.65 | 44.90% ±0.73 | 0.82 pts | Reported ±1 SE ranges overlap |
| Tax Agent Bench | 58.50% ±3.13 | 45.13% ±3.23 | 13.36 pts | Reported ±1 SE ranges do not overlap |
| TaxEval v2 | 72.08% ±0.88 | 74.65% ±0.85 | 2.58 pts | Reported ±1 SE ranges do not overlap |
| MedCode | 40.45% ±2.12 | 40.14% ±2.04 | 0.31 pts | Reported ±1 SE ranges overlap |
| MedScribe | 75.14% ±2.00 | 78.15% ±1.79 | 3.00 pts | Reported ±1 SE ranges overlap |
| GPQA Diamond | 89.39% ±1.65 | 89.14% ±2.00 | 0.25 pts | Reported ±1 SE ranges overlap |
| MMLU Pro | 87.25% ±0.34 | 87.57% ±0.33 | 0.32 pts | Reported ±1 SE ranges overlap |
| Code Migration | 26.20% ±4.04 | 27.77% ±4.14 | 1.57 pts | Reported ±1 SE ranges overlap |
| LiveCodeBench | 87.48% ±0.95 | 86.77% ±0.97 | 0.71 pts | Reported ±1 SE ranges overlap |
| ProgramBench | 0.00% ±0.00 | 0.00% ±0.00 | 0.00 pts | Reported ±1 SE ranges overlap |
| SWE-bench | 77.40% ±1.87 | 76.20% ±1.91 | 1.20 pts | Reported ±1 SE ranges overlap |
| Vibe Code Bench v1.1 | 49.93% ±4.77 | 37.89% ±4.91 | 12.04 pts | Reported ±1 SE ranges do not overlap |
| Public Benefits Bench v1.1 | 62.92% ±1.26 | 56.63% ±1.29 | 6.29 pts | Reported ±1 SE ranges do not overlap |
Performance by category
| Category | DeepSeek V4 average | Kimi K2.6 average |
|---|---|---|
| Legal | 35.72% | 34.09% |
| Finance | 56.57% | 55.64% |
| Healthcare | 57.80% | 59.15% |
| Academic | 88.32% | 88.36% |
| Coding | 48.20% | 45.73% |
| Social Mobility | 62.92% | 56.63% |
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 cost | Kimi K2.6 cost | DeepSeek V4 latency | Kimi K2.6 latency |
|---|---|---|---|---|
| Harvey's Legal Agent Benchmark | $0.74 | $0.82 | 6m42s | 23m23s |
| Legal Research Bench | $0.68 | $1.52 | 17m05s | 30m53s |
| LegalBench | N/A | N/A | 3m04s | 50.07s |
| EMB | $0.80 | $2.20 | 13m42s | 39m48s |
| Finance Agent (v2) | $0.88 | $1.14 | 25m26s | 27m10s |
| Tax Agent Bench | $0.72 | $0.57 | 20m45s | 15m24s |
| TaxEval v2 | N/A | N/A | 3m10s | 4m27s |
| MedCode | N/A | N/A | 6m23s | 5m05s |
| MedScribe | N/A | N/A | 5m46s | 8m14s |
| GPQA Diamond | N/A | N/A | 15m00s | 9m05s |
| MMLU Pro | N/A | N/A | 9m38s | 2m36s |
| Code Migration | $1.07 | $5.11 | 28m58s | 1h30m |
| LiveCodeBench | N/A | N/A | 11m56s | 8m49s |
| ProgramBench | N/A | N/A | 1h03m | 1h20m |
| SWE-bench | $0.44 | $0.27 | 10m35s | 8m40s |
| Vibe Code Bench v1.1 | $2.21 | $1.93 | 56m59s | 49m28s |
| Public Benefits Bench v1.1 | $0.46 | $0.91 | 15m22s | 9m55s |
Results available only for DeepSeek V4
- Vals Index
- ProofBench v1.1
- SkillsBench
- Terminal-Bench 4.0
Results available only for Kimi K2.6
- MortgageTax
- MMMU Pro
- SAGE