Unverified50% confidenceBenchmarkExact time
On the Terminal-Bench 2.0 benchmark, the same model GPT-5.5 scores from 66.07% to 88.09% under different harnesses (a gap of about 22 percentage points).
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
10/24/2026
First Seen
Valid until: 1/22/2027
Sources
Related Entities
Cite This Claim
Stable URI
https://kongchang.com/claim/1094309API
curl https://kongchang.com/api/v1/knowledge/claims/1094309MCP
get_claim(id=1094309)