GPT-5.6 Terra (Max): benchmark results
GPT-5.6 Terra evaluated at the max reasoning-effort setting. Provider: OpenAI. Released 2026-07-09. Access: API.
Aggregate Rating 1844 ± 8, rank #75 of 2168 rated models, from 273 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Aikido CVE Rediscovery | 83.3 | Recall (%) | 100 |
| Nejumi 4 - ARC-AGI-2 | 100 | Accuracy (%) | 100 |
| Horangi 4 - Ko-ARC-AGI | 97 | Accuracy (%) | 99.5 |
| Nejumi 4 - jaster (2-shot) - JBLiMP | 97 | Exact match (%) | 98.3 |
| Nejumi 4 - jaster (2-shot) - JCoLA (out-of-domain) | 90 | Exact match (%) | 98.3 |
| Nejumi 4 - HLE (Japanese) - Accuracy | 47.42 | Accuracy (%) | 97.9 |
| Nejumi 4 - jaster (2-shot) - ALT Japanese-to-English | 90.53 | COMET wmt22 (x100) | 97.9 |
| AA GDP.pdf - Legal - Criterion Pass Rate | 86 | Criterion Pass Rate (%) | 97.8 |
| BALROG BabaIsAI (LLM) | 97.5 | Progress (%) | 97.4 |
| AA CritPt | 30 | Accuracy (%) | 97.3 |
| AA Terminal-Bench Hard | 57.58 | Accuracy (%) | 97.3 |
| Nejumi 4 - jaster (2-shot) - NIILC | 61.45 | Character F1 (x100) | 97.2 |
Interactive version: theaggregate.ai/model?slug=gpt-5-6-terra-max · How It Works · Data refreshed daily, snapshot 2026-10-20.