Grok 4.6 vs GPT-5.6 Luna: Benchmark Comparison
Grok 4.6 has the higher score on 17 of 28 shared benchmarks; GPT-5.6 Luna leads on 11.
The largest observed score gap is 16.22 pts on MysteryMechanism , where Grok 4.6 leads.
Reported ±1 standard-error ranges overlap on 9 of 28 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 | Grok 4.6 | GPT-5.6 Luna | Gap | Reported uncertainty |
|---|---|---|---|---|
| Vals Index | 52.09% ±1.15 | 51.69% ±1.06 | 0.41 pts | Reported ±1 SE ranges overlap |
| Harvey's Legal Agent Benchmark | 15.83% ±2.65 | 1.25% ±0.83 | 14.58 pts | Reported ±1 SE ranges do not overlap |
| Legal Research Bench | 48.08% ±3.47 | 36.54% ±3.35 | 11.54 pts | Reported ±1 SE ranges do not overlap |
| LegalBench | 86.31% ±0.42 | 84.03% ±0.42 | 2.27 pts | Reported ±1 SE ranges do not overlap |
| EMB | 62.73% ±3.08 | 67.12% ±2.94 | 4.39 pts | Reported ±1 SE ranges overlap |
| Finance Agent (v2) | 53.68% ±0.67 | 55.04% ±0.31 | 1.36 pts | Reported ±1 SE ranges do not overlap |
| MortgageTax | 64.19% ±0.95 | 67.29% ±0.92 | 3.10 pts | Reported ±1 SE ranges do not overlap |
| Tax Agent Bench | 70.79% ±3.00 | 60.81% ±3.26 | 9.97 pts | Reported ±1 SE ranges do not overlap |
| TaxEval v2 | 71.10% ±0.90 | 76.17% ±0.84 | 5.07 pts | Reported ±1 SE ranges do not overlap |
| MedCode | 44.71% ±2.26 | 42.39% ±2.27 | 2.32 pts | Reported ±1 SE ranges overlap |
| MedScribe | 86.53% ±1.96 | 84.39% ±2.58 | 2.14 pts | Reported ±1 SE ranges overlap |
| ProofBench v1.1 | 51.00% ±5.02 | 60.00% ±4.92 | 9.00 pts | Reported ±1 SE ranges overlap |
| BioMysteryBench | 72.22% ±0.00 | 61.48% ±0.74 | 10.74 pts | Reported ±1 SE ranges do not overlap |
| MysteryMechanism | 30.63% ±3.10 | 14.41% ±2.36 | 16.22 pts | Reported ±1 SE ranges do not overlap |
| Terminal-Bench Science | 11.43% ±3.83 | 0.00% ±0.00 | 11.43 pts | Reported ±1 SE ranges do not overlap |
| GPQA Diamond | 94.70% ±1.13 | 91.67% ±1.74 | 3.03 pts | Reported ±1 SE ranges do not overlap |
| MMLU Pro | 89.40% ±0.30 | 86.04% ±0.35 | 3.36 pts | Reported ±1 SE ranges do not overlap |
| SAGE | 28.90% ±3.08 | 44.22% ±3.33 | 15.32 pts | Reported ±1 SE ranges do not overlap |
| Code Migration | 44.57% ±4.48 | 44.55% ±4.24 | 0.02 pts | Reported ±1 SE ranges overlap |
| IOI | 47.61% ±2.09 | 61.78% ±11.61 | 14.17 pts | Reported ±1 SE ranges do not overlap |
| ProgramBench | 1.00% ±0.70 | 0.00% ±0.00 | 1.00 pts | Reported ±1 SE ranges do not overlap |
| SkillsBench | 55.77% ±4.84 | 60.45% ±4.75 | 4.68 pts | Reported ±1 SE ranges overlap |
| SWE-bench | 95.60% ±0.92 | 93.00% ±1.14 | 2.60 pts | Reported ±1 SE ranges do not overlap |
| Terminal-Bench 4.0 | 17.17% ±1.34 | 11.62% ±1.01 | 5.56 pts | Reported ±1 SE ranges do not overlap |
| Vibe Code Bench 1-100 | 14.75% ±3.17 | 22.59% ±4.00 | 7.84 pts | Reported ±1 SE ranges do not overlap |
| Vibe Code Bench v1.1 | 76.24% ±3.82 | 77.06% ±3.07 | 0.82 pts | Reported ±1 SE ranges overlap |
| CyberBench v1.1 | 67.68% ±5.87 | 73.63% ±5.56 | 5.95 pts | Reported ±1 SE ranges overlap |
| Public Benefits Bench v1.1 | 66.85% ±1.23 | 61.16% ±1.27 | 5.68 pts | Reported ±1 SE ranges do not overlap |
Performance by category
| Category | Grok 4.6 average | GPT-5.6 Luna average |
|---|---|---|
| Legal | 50.07% | 40.61% |
| Finance | 64.50% | 65.29% |
| Healthcare | 65.62% | 63.39% |
| Math | 51.00% | 60.00% |
| Science | 38.09% | 25.30% |
| Academic | 92.05% | 88.85% |
| Education | 28.90% | 44.22% |
| Coding | 44.09% | 46.38% |
| Cyber | 67.68% | 73.63% |
| Social Mobility | 66.85% | 61.16% |
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 | Grok 4.6 cost | GPT-5.6 Luna cost | Grok 4.6 latency | GPT-5.6 Luna latency |
|---|---|---|---|---|
| Vals Index | $4.49 | $0.82 | 37m01s | 29m27s |
| Harvey's Legal Agent Benchmark | $4.01 | $0.38 | 45m05s | 10m56s |
| Legal Research Bench | $1.53 | $0.85 | 25m04s | 39m34s |
| LegalBench | N/A | N/A | 27.42s | 5.89s |
| EMB | $3.06 | $0.37 | 52m59s | 11m00s |
| Finance Agent (v2) | $1.66 | $0.28 | 16m41s | 12m52s |
| MortgageTax | N/A | N/A | 8.55s | 22.84s |
| Tax Agent Bench | $0.98 | $0.47 | 10m58s | 42m46s |
| TaxEval v2 | N/A | N/A | 47.57s | 76.52s |
| MedCode | N/A | N/A | 2m07s | 81.28s |
| MedScribe | N/A | N/A | 70.94s | 116.89s |
| ProofBench v1.1 | $0.76 | $0.12 | 14m22s | 8m34s |
| BioMysteryBench | $1.23 | $0.09 | 14m18s | 18m31s |
| MysteryMechanism | $1.31 | $0.11 | 27m55s | 8m13s |
| Terminal-Bench Science | $4.82 | $0.58 | 1h18m | 2h03m |
| GPQA Diamond | N/A | N/A | 3m29s | 53.52s |
| MMLU Pro | N/A | N/A | 54.98s | 16.95s |
| SAGE | N/A | N/A | 3m22s | 51.88s |
| Code Migration | $14.58 | $1.88 | 1h30m | 58m35s |
| IOI | $8.18 | $0.58 | 3h57m | 1h10m |
| ProgramBench | $16.76 | $0.94 | 4h27m | 41m34s |
| SkillsBench | $0.92 | $0.21 | 13m28s | 5m20s |
| SWE-bench | $0.78 | $0.04 | 10m04s | 3m21s |
| Terminal-Bench 4.0 | $5.07 | $0.73 | 29m27s | 35m55s |
| Vibe Code Bench 1-100 | $26.27 | $4.91 | 2h11m | 2h03m |
| Vibe Code Bench v1.1 | $4.88 | $0.73 | 25m30s | 25m52s |
| CyberBench v1.1 | $4.07 | $0.40 | 36m52s | 16m15s |
| Public Benefits Bench v1.1 | $0.92 | $0.27 | 25m53s | 38m40s |
Results available only for Grok 4.6
- Vals RSI Index
- LiveCodeBench
- Time Horizon Index: KSP
Results available only for GPT-5.6 Luna
- MMMU Pro