Unverified50% confidenceBenchmarkExact time
The Neural 3 base model's 4-layer version scores only 11.7 on the mobile action benchmark and 21 on droid call, while the full 20-layer version scores 86 and 47 respectively, and one developer's self-test yielded only 32.2%.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
10/5/2026
First Seen
Valid until: 1/3/2027
Sources
Related Entities
Related Claims
Unverified30%–36% of the stable neurons for instruction-tuned generators are concentrated in BERT's layer 12, whereas for base models this proportion is below 14% in all cases.63% similarUnverifiedAI-generated solutions may only match 20% or 30% of the user's ideas, and only matching 20% is essentially the same as not matching at all.62% similarUnverifiedAccording to data from Artificial Analysis, the number of reasoning tokens per task doubled from approximately 42,000 in Opus 5 to 84,000 in Opus 5.5, while output tokens only increased from 30,000 to 35,000.62% similarUnverifiedIn the experiment where the 27B model wrote its own skills, it scored 56.3% on LiveMath.61% similarUnverifiedThe developer built the ishuman benchmark, in which 2.0 was judged to be human 23.5% of the time, the base model abliterated 0.3%, the official original version 15.1%, and the base model with prompting 6.8%.61% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/977856API
curl https://kongchang.com/api/v1/knowledge/claims/977856MCP
get_claim(id=977856)