DeepSeek V4 vs MiMo V2.5 Pro: Benchmark Comparison
DeepSeek V4 has the higher score on 14 of 18 shared benchmarks; MiMo V2.5 Pro leads on 4.
The largest observed score gap is 23.71 pts on Tax Agent Bench , where DeepSeek V4 leads.
Reported ±1 standard-error ranges overlap on 6 of 18 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 | MiMo V2.5 Pro | Gap | Reported uncertainty |
|---|---|---|---|---|
| Vals Index | 38.63% ±1.16 | 33.84% ±1.17 | 4.79 pts | Reported ±1 SE ranges do not overlap |
| Harvey's Legal Agent Benchmark | 3.75% ±1.43 | 2.08% ±0.83 | 1.67 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 | 77.13% ±0.55 | 3.19 pts | Reported ±1 SE ranges do not overlap |
| EMB | 51.62% ±3.05 | 55.23% ±3.11 | 3.61 pts | Reported ±1 SE ranges overlap |
| Finance Agent (v2) | 44.08% ±0.65 | 41.50% ±1.19 | 2.58 pts | Reported ±1 SE ranges do not overlap |
| Tax Agent Bench | 58.50% ±3.13 | 34.79% ±2.62 | 23.71 pts | Reported ±1 SE ranges do not overlap |
| TaxEval v2 | 72.08% ±0.88 | 73.79% ±0.86 | 1.72 pts | Reported ±1 SE ranges overlap |
| MedCode | 40.45% ±2.12 | 32.48% ±1.91 | 7.97 pts | Reported ±1 SE ranges do not overlap |
| MedScribe | 75.14% ±2.00 | 83.73% ±2.06 | 8.59 pts | Reported ±1 SE ranges do not overlap |
| ProofBench v1.1 | 16.00% ±3.69 | 22.00% ±4.16 | 6.00 pts | Reported ±1 SE ranges overlap |
| GPQA Diamond | 89.39% ±1.65 | 82.58% ±2.47 | 6.82 pts | Reported ±1 SE ranges do not overlap |
| MMLU Pro | 87.25% ±0.34 | 84.59% ±0.45 | 2.65 pts | Reported ±1 SE ranges do not overlap |
| Code Migration | 26.20% ±4.04 | 21.56% ±4.13 | 4.64 pts | Reported ±1 SE ranges overlap |
| LiveCodeBench | 87.48% ±0.95 | 81.35% ±1.07 | 6.13 pts | Reported ±1 SE ranges do not overlap |
| SWE-bench | 77.40% ±1.87 | 74.00% ±1.96 | 3.40 pts | Reported ±1 SE ranges overlap |
| Terminal-Bench 4.0 | 11.11% ±1.34 | 0.51% ±0.51 | 10.61 pts | Reported ±1 SE ranges do not overlap |
| Vibe Code Bench v1.1 | 49.93% ±4.77 | 34.11% ±4.53 | 15.82 pts | Reported ±1 SE ranges do not overlap |
Performance by category
| Category | DeepSeek V4 average | MiMo V2.5 Pro average |
|---|---|---|
| Legal | 35.72% | 31.69% |
| Finance | 56.57% | 51.33% |
| Healthcare | 57.80% | 58.11% |
| Math | 16.00% | 22.00% |
| Academic | 88.32% | 83.59% |
| Coding | 50.43% | 42.31% |
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 | MiMo V2.5 Pro cost | DeepSeek V4 latency | MiMo V2.5 Pro latency |
|---|---|---|---|---|
| Vals Index | $1.31 | $0.16 | 32m09s | 34m33s |
| Harvey's Legal Agent Benchmark | $0.74 | $0.11 | 6m42s | 9m04s |
| Legal Research Bench | $0.68 | $0.08 | 17m05s | 6m56s |
| LegalBench | N/A | N/A | 3m04s | 27.14s |
| EMB | $0.80 | $0.22 | 13m42s | 29m08s |
| Finance Agent (v2) | $0.88 | $0.21 | 25m26s | 8m17s |
| Tax Agent Bench | $0.72 | $0.04 | 20m45s | 6m17s |
| TaxEval v2 | N/A | N/A | 3m10s | 62.83s |
| MedCode | N/A | N/A | 6m23s | 38.45s |
| MedScribe | N/A | N/A | 5m46s | 90.71s |
| ProofBench v1.1 | $0.03 | $0.09 | 7m03s | 19m37s |
| GPQA Diamond | N/A | N/A | 15m00s | 4m11s |
| MMLU Pro | N/A | N/A | 9m38s | 67.76s |
| Code Migration | $1.07 | $0.24 | 28m58s | 48m53s |
| LiveCodeBench | N/A | N/A | 11m56s | 4m31s |
| SWE-bench | $0.44 | $0.02 | 10m35s | 7m51s |
| Terminal-Bench 4.0 | $3.38 | $0.23 | 1h28m | 2h07m |
| Vibe Code Bench v1.1 | $2.21 | $0.17 | 56m59s | 40m04s |
Results available only for DeepSeek V4
- ProgramBench
- SkillsBench
- Public Benefits Bench v1.1
Results available only for MiMo V2.5 Pro
None.