Grok 4.6: benchmark results
xAI's Grok 4.6 (August 12, 2026), a flagship model for long-running agentic and visual tasks, with a 500,000-token context and a February 2026 knowledge cutoff. Provider: xAI. Released 2026-08-12. Access: API.
Aggregate Rating 1864 ± 5, rank #26 of 1656 rated models, from 593 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| AI for Education Pedagogy - Science | 96.72 | Accuracy (%) | 100 |
| Benchmarks.bio - BioSecBench-Refusal | 64.2 | Pass Rate (%) | 100 |
| Benchmarks.bio - SpatialBench | 77.39 | Pass Rate (%) | 100 |
| ExplorationBench - AlienLogic | 83.8 | Held-out accuracy after exploration (%; Best@3: the highest | 100 |
| FlavourBench | 65.07 | FlavourBench Score (0-100) | 100 |
| FlavourBench - Constraint | 58.21 | Constraint Family Score (0-100) | 100 |
| FrontierChallenge (Claude Code) | 20.6 | Pass Rate (%) | 100 |
| IndicBankBench - Bad Tool Response | 69.44 | Strict pass^3 (%; 36 cases passed out of the 36 cases of the | 100 |
| IndicBankBench - Not Enough Info | 62.3 | Strict pass^3 (%; 61 cases passed out of the 61 cases of the | 100 |
| IndicBankBench - Out-of-Scope Refusal | 77.5 | Strict pass^3 (%; 40 cases passed out of the 40 cases of the | 100 |
| Insurance Agent Benchmark (Model Alone) | 77.6 | Judge Accuracy (%) | 100 |
| MERA v2 - Agentic | 85.5 | Score (%) | 100 |
Interactive version: theaggregate.ai/model?slug=grok-4-6 · How It Works · Data refreshed daily, snapshot 2026-10-20.