Up to four models
Grok 4.6 vs Grok 4.7
On the index (GPT-4 = 100), Grok 4.6 scores 171 and Grok 4.7 168. Side by side on the same third-party tests, with price and context. The link in your address bar shares this exact comparison.
Grok 4.6
171155–188
WeightsClosed
ReleasedAug 2026
Price in / out$2.0 / $6.0
Context500K
Core tests taken6 of 9
Grok 4.7
168151–184
WeightsClosed
ReleasedSep 2026
Price in / out$2.0 / $6.0
Context500K
Core tests taken6 of 9
Grok 4.6 leads Grok 4.7 by 4 points, ahead on 5 of 6 shared core tests, at a similar price.
Test by testthird-party results
Grok 4.6Grok 4.7
Every result as a table
| Test | Grok 4.6 | Grok 4.7 |
|---|---|---|
| Reasoning & knowledge | ||
| GPQA Diamond (Epoch AI) | 94.0% | 92.7% |
| SimpleQA Verified | 49.3% | 56.0% |
| Maths | ||
| FrontierMath (tiers 1-3) | 66.0% | 53.0% |
| AIME-style maths (OTIS mock) | 99.2% | 98.1% |
| ProofBench v1.1 | 51.0% | 26.0% |
| Chess puzzles | 33.0% | 41.0% |
| FrontierMath (tier 4) | 31.7% | 17.1% |
| Coding | ||
| Terminal-Bench 4.0 | 17.2% | 28.8% |
| Code Migration | 44.6% | 44.8% |
| CyberBench v1.1 | 67.7% | 69.5% |
| IOI | 47.6% | 57.7% |
| ProgramBench | 1.0% | 0.5% |
| Vibe Code Bench v1.1 | 76.2% | 86.2% |
| FrontierSWE | 25.3% | 29.5% |
| Agents | ||
| APEX-Agents | 65.3% | 54.6% |
| BioMysteryBench | 72.2% | 69.3% |
| Terminal-Bench-Science | 11.4% | 10.0% |
| DTBench | 97.3% | 96.0% |
| Furniture assembly | 40.0% | 20.8% |
| LMCA | 48.4% | 49.4% |
| Mystery game puzzles | 34.0% | 29.0% |
| Professional work | ||
| Finance Agent (v2) | 53.7% | 52.3% |
| EMB | 62.7% | 67.0% |
| LegalBench | 86.3% | 84.4% |
| Legal Research Bench | 48.1% | 47.1% |
| MedCode | 44.7% | 49.5% |
| MedScribe | 86.5% | 89.4% |
| Public Benefits Bench v1.1 | 66.8% | 65.6% |
| SAGE | 28.9% | 40.8% |
| Tax Agent Bench | 33.1% | 34.2% |
| Novel problems | ||
| ARC-AGI-2 | 67.1% | 61.4% |
| ARC-AGI-1 | 87.5% | 90.2% |
| ARC-AGI-3 | 2.1% | 1.8% |
| MysteryMechanism | 30.6% | 25.2% |
Tests at least two of these models have taken. Bold rows are the core tests.
