Skip to main content
leaderboard
.cn
Toggle Sidebar
Search models or organizations…
⌘K
Explore
Frontier Models
Historical Trends
Local Models
Compare Models
Ranking Studio
Library
Models
Pricing
Organizations
Benchmarks
Data sources
Settings
Toggle Sidebar
leaderboard
.cn
Loading page content
Feedback
Benchmark details loading
Back to benchmarks
Factual error rate on difficult prompts (2026-10-26)
Reasoning
Share
Official website
Response-level factual error rate on the refreshed October 26 difficult-prompt evaluation; not directly comparable to earlier GPT-6 system cards or representative of production usage.
Ranking
Sources
#
Model
Score
1
GPT-6 Astra
OpenAI · max
4.2
%
2
GPT-6.1 Terra
OpenAI · max
4.3
%
2 models total