MiMo V2.5 vs GLM 5.3 Flash: Benchmark Comparison
MiMo V2.5 has the higher score on 2 of 16 shared benchmarks; GLM 5.3 Flash leads on 14.
The largest observed score gap is 36.06 pts on Legal Research Bench , where GLM 5.3 Flash leads.
Reported ±1 standard-error ranges overlap on 4 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 | MiMo V2.5 | GLM 5.3 Flash | Gap | Reported uncertainty |
|---|---|---|---|---|
| Harvey's Legal Agent Benchmark | 1.67% ±0.00 | 6.67% ±1.65 | 5.00 pts | Reported ±1 SE ranges do not overlap |
| Legal Research Bench | 9.13% ±2.00 | 45.19% ±3.46 | 36.06 pts | Reported ±1 SE ranges do not overlap |
| LegalBench | 78.89% ±0.51 | 83.93% ±0.43 | 5.04 pts | Reported ±1 SE ranges do not overlap |
| EMB | 55.09% ±3.13 | 55.93% ±3.56 | 0.84 pts | Reported ±1 SE ranges overlap |
| Finance Agent (v2) | 36.73% ±0.11 | 57.85% ±2.06 | 21.12 pts | Reported ±1 SE ranges do not overlap |
| Tax Agent Bench | 29.16% ±2.71 | 61.96% ±3.26 | 32.79 pts | Reported ±1 SE ranges do not overlap |
| TaxEval v2 | 71.83% ±0.88 | 75.59% ±0.84 | 3.76 pts | Reported ±1 SE ranges do not overlap |
| MedScribe | 72.15% ±1.85 | 88.94% ±1.91 | 16.79 pts | Reported ±1 SE ranges do not overlap |
| ProofBench v1.1 | 16.00% ±3.69 | 21.00% ±4.09 | 5.00 pts | Reported ±1 SE ranges overlap |
| GPQA Diamond | 81.57% ±2.05 | 86.36% ±2.13 | 4.80 pts | Reported ±1 SE ranges do not overlap |
| MMLU Pro | 82.93% ±0.37 | 86.06% ±0.34 | 3.13 pts | Reported ±1 SE ranges do not overlap |
| MMMU Pro | 80.00% ±0.96 | 86.01% ±0.83 | 6.01 pts | Reported ±1 SE ranges do not overlap |
| Code Migration | 14.23% ±3.69 | 20.52% ±4.21 | 6.29 pts | Reported ±1 SE ranges overlap |
| LiveCodeBench | 81.51% ±1.07 | 80.51% ±1.05 | 1.00 pts | Reported ±1 SE ranges overlap |
| SWE-bench | 71.00% ±2.03 | 92.00% ±1.21 | 21.00 pts | Reported ±1 SE ranges do not overlap |
| Vibe Code Bench v1.1 | 42.17% ±4.57 | 30.76% ±5.28 | 11.41 pts | Reported ±1 SE ranges do not overlap |
Performance by category
| Category | MiMo V2.5 average | GLM 5.3 Flash average |
|---|---|---|
| Legal | 29.90% | 45.26% |
| Finance | 48.20% | 62.83% |
| Healthcare | 72.15% | 88.94% |
| Math | 16.00% | 21.00% |
| Academic | 81.50% | 86.14% |
| Coding | 52.23% | 55.95% |
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 | MiMo V2.5 cost | GLM 5.3 Flash cost | MiMo V2.5 latency | GLM 5.3 Flash latency |
|---|---|---|---|---|
| Harvey's Legal Agent Benchmark | $0.04 | $0.57 | 6m59s | 52m17s |
| Legal Research Bench | $0.03 | $0.12 | 6m31s | 51m07s |
| LegalBench | N/A | N/A | 11.18s | 12.52s |
| EMB | $0.08 | $0.71 | 30m38s | 1h15m |
| Finance Agent (v2) | $0.09 | $0.05 | 5m39s | 14m51s |
| Tax Agent Bench | $0.02 | $0.14 | 8m21s | 21m37s |
| TaxEval v2 | N/A | N/A | 26.30s | 53.06s |
| MedScribe | N/A | N/A | 20.26s | 86.98s |
| ProofBench v1.1 | $0.06 | $0.22 | 20m56s | 1h11m |
| GPQA Diamond | N/A | N/A | 89.39s | 2m11s |
| MMLU Pro | N/A | N/A | 30.25s | 44.76s |
| MMMU Pro | N/A | N/A | 42.47s | 28.22s |
| Code Migration | $0.10 | $3.13 | 37m55s | 3h26m |
| LiveCodeBench | N/A | N/A | 108.61s | 5m02s |
| SWE-bench | $0.01 | $0.02 | 4m04s | 1h04m |
| Vibe Code Bench v1.1 | $0.07 | $2.35 | 26m20s | 1h24m |
Results available only for MiMo V2.5
- MortgageTax
- MedCode
- SAGE
Results available only for GLM 5.3 Flash
- Terminal-Bench Science
- IOI
- ProgramBench
- SkillsBench
- Vibe Code Bench 1-100
- CyberBench v1.1