Unverified50% confidenceBenchmarkExact time
In eight benchmarks run by Exa itself, Exa 3.0 led Perplexity, Brave and Parallel, with the largest gap on BrowseComp at 91.8%.
1
Sources
50%
Confidence
Short-term (~14 days)
Relevance
10/20/2026
First Seen
Valid until: 11/3/2026
Sources
Devin Now Commits 89% of Cognition's Code as Company Hits $1B Annualized Revenue
youtubeNerra Network10/20/2026
Related Entities
Related Claims
UnverifiedDeepSeek V4.1 claims to achieve 74 on SWE, but the blogger found that this was the best result cherry-picked from multiple tests, with open code testing at 65 and PI testing at 66.56% similarUnverifiedThe intelligent routing solution passes 9/12 (75%) tasks and Opus passes 11/12 (92%); of the three failures, two unit test tasks and one QA task were routed to Claude Haiku 5.5, whose responses exceeded the 8,000-token limit.56% similarUnverifiedIn coding tests, Opus 5.5 performs best at the medium reasoning level, with score drops occurring at the high and ultra-high levels.55% similarUnverifiedIn the AID benchmark, o1 high (corresponding to o1 pro) scored above 60 in programming ability, outperforming o3-mini-high, though the two are very close.55% similarUnverifiedEmsyAI was trained for 15,000 steps using FP16 mixed precision, with validation perplexity converging to 5.8654% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1074810API
curl https://kongchang.com/api/v1/knowledge/claims/1074810MCP
get_claim(id=1074810)