Unverified50% confidenceBenchmarkExact time
Haiku 5.5's terminal programming Terminal-Bench score is 39.2%, which is 31 points lower than Sonnet 5.5's 70.6%, while the knowledge work GDPVal increased from 735 in the previous generation to 1620.
1
Sources
50%
Confidence
Long-term
Relevance
10/11/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedHaiku 5.5's score on Terminal Bench 4.0 rose from 0% in the previous generation to 39.2%.78% similarUnverifiedHaiku 5.5's score on computer use tasks jumped from 16% to 72%72% similarUnverifiedIn the creator's benchmark, Haiku 5.5's Terminal Bench score improved from 0% on Haiku 4.5 to nearly 40%.71% similarUnverifiedHaiku 5.5 outperforms GPT-6 Luna by approximately 40% on Computer Use tasks.70% similarUnverifiedHaiku 5.5 achieved a score of 1620 on the relevant coding benchmark, far higher than Haiku 4.5's score of 735.70% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/995475API
curl https://kongchang.com/api/v1/knowledge/claims/995475MCP
get_claim(id=995475)