Unverified50% confidenceBenchmarkExact time
In Paul Well-Hurren's bug-fixing test, Terra's highest level scored 40.7, versus 32 for the previous generation's highest level; however, Terra's ultra-high level scored only 11.3 and its high level 9.7, lower than the previous generation at the same levels.
1
Sources
50%
Confidence
Short-term (~14 days)
Relevance
11/1/2026
First Seen
Valid until: 11/15/2026
Sources
Related Entities
Related Claims
UnverifiedAccording to data published by OpenAI, on the Agents Last Exam benchmark, at medium reasoning effort, GPT-6.1 Terra scored 57.6% and GPT-5.6 Terra scored 42.6%, an improvement of 15 percentage points.68% similarUnverifiedOn the command-line coding test, Terra's highest level raised its score from 21.5% in the previous generation to 50.3%.68% similarUnverifiedOpenAI's internal high-difficulty fact test shows that 5.6 Terra's error rate at low reasoning effort was as high as 53.6%, while GPT-6.1 Terra cut the error rate to 16.7%, and further reduced it to 4.3% at maximum reasoning.66% similarUnverifiedOn the independent Intelligence Index (version 4.3), Terra scored 48, while Shou (the flagship) scored 52 and Opus 5.5 scored 58 under the same methodology.65% similarUnverified跑分排名不等于所有场景的实际表现，Terra跑分接近5.5但游戏任务中5.5完成度略高64% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1121969API
curl https://kongchang.com/api/v1/knowledge/claims/1121969MCP
get_claim(id=1121969)