Mercury 2.5 vs MiniMax-M3: Benchmark Comparison
Mercury 2.5 has the higher score on 0 of 11 shared benchmarks; MiniMax-M3 leads on 11.
The largest observed score gap is 43.88 pts on Vibe Code Bench v1.1 , where MiniMax-M3 leads.
Reported ±1 standard-error ranges overlap on 1 of 11 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 | Mercury 2.5 | MiniMax-M3 | Gap | Reported uncertainty |
|---|---|---|---|---|
| Legal Research Bench | 4.33% ±1.41 | 29.81% ±3.18 | 25.48 pts | Reported ±1 SE ranges do not overlap |
| LegalBench | 83.06% ±0.48 | 85.42% ±0.44 | 2.36 pts | Reported ±1 SE ranges do not overlap |
| Tax Agent Bench | 2.49% ±1.02 | 20.59% ±2.87 | 18.10 pts | Reported ±1 SE ranges do not overlap |
| MedCode | 31.33% ±1.95 | 46.29% ±2.10 | 14.96 pts | Reported ±1 SE ranges do not overlap |
| MedScribe | 55.09% ±2.10 | 87.25% ±1.96 | 32.16 pts | Reported ±1 SE ranges do not overlap |
| ProofBench v1.1 | 3.00% ±1.71 | 18.00% ±3.86 | 15.00 pts | Reported ±1 SE ranges do not overlap |
| SkillsBench | 18.06% ±3.08 | 51.50% ±4.50 | 33.44 pts | Reported ±1 SE ranges do not overlap |
| Terminal-Bench 4.0 | 0.00% ±0.00 | 1.01% ±1.01 | 1.01 pts | Reported ±1 SE ranges overlap |
| Vibe Code Bench v1.1 | 3.68% ±1.41 | 47.57% ±5.44 | 43.88 pts | Reported ±1 SE ranges do not overlap |
| CyberBench v1.1 | 45.48% ±6.06 | 65.00% ±6.10 | 19.52 pts | Reported ±1 SE ranges do not overlap |
| Public Benefits Bench v1.1 | 45.26% ±1.29 | 64.14% ±1.25 | 18.88 pts | Reported ±1 SE ranges do not overlap |
Performance by category
| Category | Mercury 2.5 average | MiniMax-M3 average |
|---|---|---|
| Legal | 43.69% | 57.61% |
| Finance | 2.49% | 20.59% |
| Healthcare | 43.21% | 66.77% |
| Math | 3.00% | 18.00% |
| Coding | 7.25% | 33.36% |
| Cyber | 45.48% | 65.00% |
| Social Mobility | 45.26% | 64.14% |
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 | Mercury 2.5 cost | MiniMax-M3 cost | Mercury 2.5 latency | MiniMax-M3 latency |
|---|---|---|---|---|
| Legal Research Bench | $0.06 | $0.34 | 47.61s | 13m34s |
| LegalBench | N/A | N/A | 3.91s | 8.21s |
| Tax Agent Bench | $0.02 | $0.16 | 29.67s | 5m15s |
| MedCode | N/A | N/A | 7.93s | 63.12s |
| MedScribe | N/A | N/A | 9.99s | 2m04s |
| ProofBench v1.1 | $0.04 | $0.42 | 3m22s | 11m59s |
| SkillsBench | $0.10 | $0.50 | 49.36s | 15m28s |
| Terminal-Bench 4.0 | $0.37 | $7.23 | 15m14s | 1h16m |
| Vibe Code Bench v1.1 | $0.38 | $6.45 | 3m13s | 1h25m |
| CyberBench v1.1 | $0.12 | $3.41 | 18m59s | 40m56s |
| Public Benefits Bench v1.1 | $0.07 | $0.25 | 68.33s | 11m31s |
Results available only for Mercury 2.5
- Terminal-Bench-Science
- IOI
Results available only for MiniMax-M3
- Vals Finance Index
- Vals Inference Provider Index
- EMB v1.1
- Finance Agent v2.1
- MortgageTax
- TaxEval v2
- GPQA Diamond
- MMLU Pro
- MMMU Pro
- SAGE
- Code Migration v1.1
- LiveCodeBench
- SWE-bench
- SWE-bench Multimodal
- Vibe Code Bench 1-100