Unverified50% confidenceBenchmarkExact time
On KEEP-Bench, the best open-source model (31B) retained all required evidence in only 66.8% of cases.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
10/20/2026
First Seen
Valid until: 1/18/2027
Sources
Related Entities
Related Claims
UnverifiedIn expert review, DiscoveryRSI-27B scored higher than GPT-6 Luna on two items: hypothesis importance and "whether the claim is credible and has evidence boundaries."57% similarUnverifiedFor self-hosted workloads primarily involving labeling, parsing, and simple Q&A, a high-quality 9B–14B model is entirely sufficient, and blindly pursuing a 26B model yields diminishing returns57% similarUnverifiedThe best-performing local model, llama3.1:8b, still guessed 31 out of 40 unanswerable questions, while every frontier model on Kaggle held steady at a false-confidence rate of 5% or below.57% similarUnverifiedIn DeepSeek's official comparison, when running with the same checkpoint at best effort, its minimal framework scored 72.6 while the open-source code scored 65.5.56% similarUnverifiedA Reddit technical post claims that a model with only 0.8B parameters achieved a score of 42.15% on the ARC-Challenge benchmark, surpassing the untuned stock Qwen3.5-2B base model and outperforming a solution using SFT distillation with 25,000 data entries.56% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1073300API
curl https://kongchang.com/api/v1/knowledge/claims/1073300MCP
get_claim(id=1073300)