Unverified50% confidenceBenchmarkExact time
According to OpenAI, the Terra model's score on a command-line coding benchmark is more than double that of the old version.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
10/28/2026
First Seen
Valid until: 1/26/2027
Sources
Cursor Agent SDK Released: Agents as Code, Evaluation Decides Everything
youtubeAI Developer Brief10/27/2026
Related Entities
Related Claims
UnverifiedOpenAI explained that the reason for the usage cut was that "the new model is twice as efficient."68% similarUnverifiedThe gains from increasing model parameters are approximately twice those from performing a second pass over the data.64% similarUnverifiedWhen a language model computes a numerical average, running it twice in succession may yield different results (e.g., 49.6 and 49.8), both of which can pass schema validation but may not be correct.64% similarUnverifiedThe model is trained in two stages: first masked language model pre-training on about 20 million sentences from FineWeb-Edu, then correction fine-tuning on 12.02 million (noisy, clean) pairs, of which 30% are identity pairs.63% similarUnverifiedOn two tasks, biomedical inference and vulnerability detection, tested against 18 models, the best-of-10 attack flipped correct answers for an average of 54.9% of cases on proprietary models and reached 80.0% on open-weight models.62% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1107102API
curl https://kongchang.com/api/v1/knowledge/claims/1107102MCP
get_claim(id=1107102)