Unverified50% confidenceDecision RuleExact time
Different versions of the same benchmark (such as Terminal Bench 4.0 and 4.1) have different problem sets and scoring rules, and scores cannot be compared across versions.
1
Sources
50%
Confidence
Long-term
Relevance
11/15/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedA score of about 65% on Terminal-Bench is already above average, but comparisons across different versions (such as 4.0) require caution, because the task set and scoring rules may be adjusted.69% similarUnverifiedTerminal-Bench 4.0 has a standard error of ±2.6 points, and Opus 5's 51.8% reproduced as 52.3%, which Anthropic calls noise.66% similarUnverifiedWhen independent evaluators run Terminal Bench 2.1, models such as MiMo 2.6 Pro and Qwen3 32B Max score in the 60s, far below the vendors' published values, and the evaluation framework itself is also changing, so top scores are not fully comparable.63% similarUnverifiedFable 5.1 scored 52.6% on the Terminal Bench Science 0.1 benchmark, more than doubling the score of its previous generation.61% similarUnverifiedOn the Terminal-Bench 2.0 benchmark, the same model Opus 4.6 scores from 60.45% to 82.02% under different harnesses (a gap of about 22 percentage points).60% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1172029API
curl https://kongchang.com/api/v1/knowledge/claims/1172029MCP
get_claim(id=1172029)