GPT-6 Luna vs MiMo V2.6 Pro: Benchmark Comparison
GPT-6 Luna has the higher score on 6 of 19 shared benchmarks; MiMo V2.6 Pro leads on 12.
The largest observed score gap is 17.68 pts on Terminal-Bench 4.0 , where MiMo V2.6 Pro leads.
Reported ±1 standard-error ranges overlap on 11 of 19 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 | GPT-6 Luna | MiMo V2.6 Pro | Gap | Reported uncertainty |
|---|---|---|---|---|
| Vals Index | 51.22% ±1.07 | 55.20% ±1.18 | 3.98 pts | Reported ±1 SE ranges do not overlap |
| Harvey's Legal Agent Benchmark | 2.92% ±1.36 | 10.83% ±2.45 | 7.92 pts | Reported ±1 SE ranges do not overlap |
| Legal Research Bench | 30.29% ±3.19 | 47.12% ±3.47 | 16.83 pts | Reported ±1 SE ranges do not overlap |
| EMB | 68.52% ±2.81 | 62.86% ±3.14 | 5.66 pts | Reported ±1 SE ranges overlap |
| Finance Agent (v2) | 49.87% ±0.23 | 57.34% ±0.57 | 7.47 pts | Reported ±1 SE ranges do not overlap |
| Tax Agent Bench | 58.86% ±3.21 | 64.94% ±3.21 | 6.07 pts | Reported ±1 SE ranges overlap |
| MedCode | 44.69% ±2.30 | 44.97% ±2.10 | 0.28 pts | Reported ±1 SE ranges overlap |
| MedScribe | 83.71% ±1.95 | 88.31% ±1.94 | 4.60 pts | Reported ±1 SE ranges do not overlap |
| ProofBench v1.1 | 64.00% ±4.82 | 70.00% ±4.61 | 6.00 pts | Reported ±1 SE ranges overlap |
| MysteryMechanism | 19.37% ±2.66 | 15.31% ±2.42 | 4.05 pts | Reported ±1 SE ranges overlap |
| Terminal-Bench Science | 4.29% ±2.44 | 2.86% ±2.01 | 1.43 pts | Reported ±1 SE ranges overlap |
| SAGE | 48.09% ±3.38 | 45.05% ±3.40 | 3.04 pts | Reported ±1 SE ranges overlap |
| Code Migration | 42.55% ±4.42 | 43.01% ±4.32 | 0.46 pts | Reported ±1 SE ranges overlap |
| IOI | 55.56% ±8.87 | 39.33% ±2.41 | 16.22 pts | Reported ±1 SE ranges do not overlap |
| ProgramBench | 0.50% ±0.50 | 0.50% ±0.50 | 0.00 pts | Reported ±1 SE ranges overlap |
| Terminal-Bench 4.0 | 13.64% ±1.51 | 31.31% ±3.07 | 17.68 pts | Reported ±1 SE ranges do not overlap |
| Vibe Code Bench v1.1 | 81.65% ±3.38 | 85.22% ±3.39 | 3.57 pts | Reported ±1 SE ranges overlap |
| CyberBench v1.1 | 76.25% ±5.29 | 72.86% ±5.50 | 3.39 pts | Reported ±1 SE ranges overlap |
| Public Benefits Bench v1.1 | 57.65% ±1.28 | 68.94% ±1.20 | 11.30 pts | Reported ±1 SE ranges do not overlap |
Performance by category
| Category | GPT-6 Luna average | MiMo V2.6 Pro average |
|---|---|---|
| Legal | 16.60% | 28.97% |
| Finance | 59.08% | 61.71% |
| Healthcare | 64.20% | 66.64% |
| Math | 64.00% | 70.00% |
| Science | 11.83% | 9.09% |
| Education | 48.09% | 45.05% |
| Coding | 38.78% | 39.88% |
| Cyber | 76.25% | 72.86% |
| Social Mobility | 57.65% | 68.94% |
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 | GPT-6 Luna cost | MiMo V2.6 Pro cost | GPT-6 Luna latency | MiMo V2.6 Pro latency |
|---|---|---|---|---|
| Vals Index | $0.43 | $0.41 | 30m34s | 1h07m |
| Harvey's Legal Agent Benchmark | $0.30 | $0.22 | 17m32s | 25m49s |
| Legal Research Bench | $0.44 | $0.18 | 42m28s | 30m16s |
| EMB | $0.14 | $0.33 | 17m21s | 55m14s |
| Finance Agent (v2) | $0.12 | $0.20 | 14m58s | 10m48s |
| Tax Agent Bench | $0.19 | $0.13 | 33m16s | 25m48s |
| MedCode | N/A | N/A | 108.77s | 2m57s |
| MedScribe | N/A | N/A | 2m50s | 2m53s |
| ProofBench v1.1 | $0.04 | $0.25 | 7m31s | 1h02m |
| MysteryMechanism | $0.03 | $0.15 | 6m01s | 44m00s |
| Terminal-Bench Science | $0.24 | $0.62 | 1h16m | 3h12m |
| SAGE | N/A | N/A | 91.07s | 3m40s |
| Code Migration | $0.60 | $0.69 | 50m23s | 2h49m |
| IOI | $0.20 | $0.84 | 35m47s | 2h26m |
| ProgramBench | $0.18 | $0.73 | 28m55s | 3h00m |
| Terminal-Bench 4.0 | $0.35 | $0.50 | 32m46s | 2h42m |
| Vibe Code Bench v1.1 | $1.35 | $1.04 | 37m00s | 1h01m |
| CyberBench v1.1 | $0.13 | $0.09 | 15m50s | 27m41s |
| Public Benefits Bench v1.1 | $0.22 | $0.08 | 52m59s | 37m14s |
Results available only for GPT-6 Luna
- BioMysteryBench
Results available only for MiMo V2.6 Pro
- SRE Bench