Kimi K2.6 vs GPT 5.4 Mini: Benchmark Comparison
Kimi K2.6 has the higher score on 12 of 16 shared benchmarks; GPT 5.4 Mini leads on 3.
The largest observed score gap is 14.83 pts on Code Migration , where Kimi K2.6 leads.
Reported ±1 standard-error ranges overlap on 6 of 16 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 K2.6 | GPT 5.4 Mini | Gap | Reported uncertainty |
|---|---|---|---|---|
| Harvey's Legal Agent Benchmark | 1.67% ±0.83 | 0.00% ±0.00 | 1.67 pts | Reported ±1 SE ranges do not overlap |
| Legal Research Bench | 15.87% ±2.54 | 12.50% ±2.30 | 3.37 pts | Reported ±1 SE ranges overlap |
| EMB | 57.85% ±2.74 | 45.43% ±3.65 | 12.43 pts | Reported ±1 SE ranges do not overlap |
| Finance Agent (v2) | 44.90% ±0.73 | 45.36% ±0.45 | 0.46 pts | Reported ±1 SE ranges overlap |
| MortgageTax | 65.82% ±0.93 | 63.51% ±0.91 | 2.30 pts | Reported ±1 SE ranges do not overlap |
| Tax Agent Bench | 45.13% ±3.23 | 36.59% ±3.10 | 8.55 pts | Reported ±1 SE ranges do not overlap |
| TaxEval v2 | 74.65% ±0.85 | 71.22% ±0.90 | 3.43 pts | Reported ±1 SE ranges do not overlap |
| GPQA Diamond | 89.14% ±2.00 | 83.08% ±2.46 | 6.06 pts | Reported ±1 SE ranges do not overlap |
| MMLU Pro | 87.57% ±0.33 | 84.55% ±0.36 | 3.02 pts | Reported ±1 SE ranges do not overlap |
| MMMU Pro | 86.30% ±0.83 | 79.25% ±0.97 | 7.05 pts | Reported ±1 SE ranges do not overlap |
| SAGE | 50.22% ±3.43 | 50.81% ±3.40 | 0.59 pts | Reported ±1 SE ranges overlap |
| Code Migration | 27.77% ±4.14 | 12.94% ±3.64 | 14.83 pts | Reported ±1 SE ranges do not overlap |
| LiveCodeBench | 86.77% ±0.97 | 81.47% ±1.09 | 5.31 pts | Reported ±1 SE ranges do not overlap |
| ProgramBench | 0.00% ±0.00 | 0.00% ±0.00 | 0.00 pts | Reported ±1 SE ranges overlap |
| SWE-bench | 76.20% ±1.91 | 73.00% ±1.99 | 3.20 pts | Reported ±1 SE ranges overlap |
| Vibe Code Bench v1.1 | 37.89% ±4.91 | 47.97% ±5.61 | 10.08 pts | Reported ±1 SE ranges overlap |
Performance by category
| Category | Kimi K2.6 average | GPT 5.4 Mini average |
|---|---|---|
| Legal | 8.77% | 6.25% |
| Finance | 57.67% | 52.42% |
| Academic | 87.67% | 82.29% |
| Education | 50.22% | 50.81% |
| Coding | 45.73% | 43.07% |
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 K2.6 cost | GPT 5.4 Mini cost | Kimi K2.6 latency | GPT 5.4 Mini latency |
|---|---|---|---|---|
| Harvey's Legal Agent Benchmark | $0.82 | $1.00 | 23m23s | 11m21s |
| Legal Research Bench | $1.52 | $1.86 | 30m53s | 37m36s |
| EMB | $2.20 | $2.15 | 39m48s | 43m21s |
| Finance Agent (v2) | $1.14 | $1.20 | 27m10s | 35m54s |
| MortgageTax | N/A | N/A | 109.76s | 2m06s |
| Tax Agent Bench | $0.57 | $0.81 | 15m24s | 16m46s |
| TaxEval v2 | N/A | N/A | 4m27s | 44.58s |
| GPQA Diamond | N/A | N/A | 9m05s | 63.11s |
| MMLU Pro | N/A | N/A | 2m36s | 20.00s |
| MMMU Pro | N/A | N/A | 4m02s | 50.81s |
| SAGE | N/A | N/A | 4m31s | 2m21s |
| Code Migration | $5.11 | $1.95 | 1h30m | 28m53s |
| LiveCodeBench | N/A | N/A | 8m49s | 3m09s |
| ProgramBench | N/A | N/A | 1h20m | 53m45s |
| SWE-bench | $0.27 | $0.51 | 8m40s | 5m26s |
| Vibe Code Bench v1.1 | $1.93 | $1.19 | 49m28s | 34m15s |
Results available only for Kimi K2.6
- LegalBench
- MedCode
- MedScribe
- Public Benefits Bench v1.1
Results available only for GPT 5.4 Mini
- Vals Index
- Terminal-Bench 4.0