GPT-5.6 Terra: benchmark results
OpenAI's balanced GPT-5.6 tier for everyday reasoning and coding work. Provider: OpenAI. Released 2026-07-09. Access: API.
Aggregate Rating 1836 ± 4, rank #40 of 1635 rated models, from 998 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| AgentIdeaBench - Static - Feasibility | 6.53 | Feasibility Critic Score (1-10) | 100 |
| Aikido CVE Rediscovery | 83.3 | Recall (%) | 100 |
| ClashBench - Deliberate Interference Rate (Codex) | 43.2 | DIR (%) | 100 |
| ClashBench - Successful Preemption Rate (Codex) | 34.5 | SPR (%) | 100 |
| IFM K2-Horizon-375B Card - MCPMark | 74 | Score (%) | 100 |
| IFM K2-Horizon-375B Card - WildClawBench (English Text-Only Subset) | 60 | Score (%) | 100 |
| Nejumi 4 - ARC-AGI-2 | 100 | Accuracy (%) | 100 |
| Nejumi 4 - GLP - Abstract Reasoning | 98 | Score (%) | 100 |
| OpenAI GPT-5.6 Launch - NanoGPT | 14.5 | Normalized reward (%) | 100 |
| OpenAI GPT-5.6 Launch - PostTrainBench Lite | 51.5 | Score (%) | 100 |
| OpenAI GPT-5.6 System Card - Challenging Prompts - Sexual | 96.6 | not_unsafe rate (%) | 100 |
| OpenAI GPT-5.6 System Card - Cyber Safety - Production Data | 0.99 | Safety score (0-1) | 100 |
Interactive version: theaggregate.ai/model?slug=gpt-5-6-terra · How It Works · Data refreshed daily, snapshot 2026-10-15.