Grok 4.7 vs GPT-5.6 Luna: Benchmark Comparison
Grok 4.7 has the higher score on 15 of 21 shared benchmarks; GPT-5.6 Luna leads on 6.
The largest observed score gap is 34.00 pts on ProofBench v1.1 , where GPT-5.6 Luna leads.
Reported ±1 standard-error ranges overlap on 8 of 21 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 4.7 | GPT-5.6 Luna | Gap | Reported uncertainty |
|---|---|---|---|---|
| Vals Index | 54.95% ±1.07 | 51.69% ±1.06 | 3.26 pts | Reported ±1 SE ranges do not overlap |
| Harvey's Legal Agent Benchmark | 12.50% ±2.66 | 1.25% ±0.83 | 11.25 pts | Reported ±1 SE ranges do not overlap |
| Legal Research Bench | 47.12% ±3.47 | 36.54% ±3.35 | 10.58 pts | Reported ±1 SE ranges do not overlap |
| LegalBench | 84.39% ±0.46 | 84.03% ±0.42 | 0.35 pts | Reported ±1 SE ranges overlap |
| EMB | 66.99% ±3.04 | 67.12% ±2.94 | 0.13 pts | Reported ±1 SE ranges overlap |
| Finance Agent (v2) | 52.25% ±0.35 | 55.04% ±0.31 | 2.79 pts | Reported ±1 SE ranges do not overlap |
| Tax Agent Bench | 65.60% ±3.24 | 60.81% ±3.26 | 4.78 pts | Reported ±1 SE ranges overlap |
| MedCode | 49.55% ±2.17 | 42.39% ±2.27 | 7.16 pts | Reported ±1 SE ranges do not overlap |
| MedScribe | 89.38% ±1.89 | 84.39% ±2.58 | 4.99 pts | Reported ±1 SE ranges do not overlap |
| ProofBench v1.1 | 26.00% ±4.41 | 60.00% ±4.92 | 34.00 pts | Reported ±1 SE ranges do not overlap |
| BioMysteryBench | 69.26% ±1.33 | 61.48% ±0.74 | 7.78 pts | Reported ±1 SE ranges do not overlap |
| MysteryMechanism | 25.23% ±2.92 | 14.41% ±2.36 | 10.81 pts | Reported ±1 SE ranges do not overlap |
| Terminal-Bench Science | 10.00% ±3.61 | 0.00% ±0.00 | 10.00 pts | Reported ±1 SE ranges do not overlap |
| SAGE | 40.79% ±3.35 | 44.22% ±3.33 | 3.43 pts | Reported ±1 SE ranges overlap |
| Code Migration | 44.82% ±4.21 | 44.55% ±4.24 | 0.27 pts | Reported ±1 SE ranges overlap |
| IOI | 57.72% ±1.99 | 61.78% ±11.61 | 4.06 pts | Reported ±1 SE ranges overlap |
| ProgramBench | 0.50% ±0.50 | 0.00% ±0.00 | 0.50 pts | Reported ±1 SE ranges overlap |
| Terminal-Bench 4.0 | 28.79% ±2.31 | 11.62% ±1.01 | 17.17 pts | Reported ±1 SE ranges do not overlap |
| Vibe Code Bench v1.1 | 86.17% ±2.18 | 77.06% ±3.07 | 9.12 pts | Reported ±1 SE ranges do not overlap |
| CyberBench v1.1 | 69.46% ±5.67 | 73.63% ±5.56 | 4.17 pts | Reported ±1 SE ranges overlap |
| Public Benefits Bench v1.1 | 65.63% ±1.24 | 61.16% ±1.27 | 4.47 pts | Reported ±1 SE ranges do not overlap |
Performance by category
| Category | Grok 4.7 average | GPT-5.6 Luna average |
|---|---|---|
| Legal | 48.00% | 40.61% |
| Finance | 61.61% | 60.99% |
| Healthcare | 69.47% | 63.39% |
| Math | 26.00% | 60.00% |
| Science | 34.83% | 25.30% |
| Education | 40.79% | 44.22% |
| Coding | 43.60% | 39.00% |
| Cyber | 69.46% | 73.63% |
| Social Mobility | 65.63% | 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 4.7 cost | GPT-5.6 Luna cost | Grok 4.7 latency | GPT-5.6 Luna latency |
|---|---|---|---|---|
| Vals Index | $12.12 | $0.82 | 35m31s | 29m27s |
| Harvey's Legal Agent Benchmark | $11.13 | $0.38 | 42m05s | 10m56s |
| Legal Research Bench | $4.92 | $0.85 | 21m14s | 39m34s |
| LegalBench | N/A | N/A | 26.02s | 5.89s |
| EMB | $6.48 | $0.37 | 34m57s | 11m00s |
| Finance Agent (v2) | $2.72 | $0.28 | 15m33s | 12m52s |
| Tax Agent Bench | $1.80 | $0.47 | 16m36s | 42m46s |
| MedCode | N/A | N/A | 3m33s | 81.28s |
| MedScribe | N/A | N/A | 2m33s | 116.89s |
| ProofBench v1.1 | $0.79 | $0.12 | 12m47s | 8m34s |
| BioMysteryBench | $2.49 | $0.09 | 11m37s | 18m31s |
| MysteryMechanism | $1.84 | $0.11 | 17m33s | 8m13s |
| Terminal-Bench Science | $14.45 | $0.58 | 1h29m | 2h03m |
| SAGE | N/A | N/A | 4m51s | 51.88s |
| Code Migration | $36.55 | $1.88 | 1h09m | 58m35s |
| IOI | $12.71 | $0.58 | 46m48s | 1h10m |
| ProgramBench | $46.49 | $0.94 | 3h38m | 41m34s |
| Terminal-Bench 4.0 | $18.09 | $0.73 | 50m21s | 35m55s |
| Vibe Code Bench v1.1 | $15.83 | $0.73 | 35m47s | 25m52s |
| CyberBench v1.1 | $8.54 | $0.40 | 15m30s | 16m15s |
| Public Benefits Bench v1.1 | $2.25 | $0.27 | 16m57s | 38m40s |
Results available only for Grok 4.7
- Vals RSI Index
Results available only for GPT-5.6 Luna
- MortgageTax
- TaxEval v2
- GPQA Diamond
- MMLU Pro
- MMMU Pro
- SkillsBench
- SWE-bench
- Vibe Code Bench 1-100