Unverified50% confidenceBenchmarkExact time
In C++ language tests, Swift's token usage was only 29% of the base version's, but accounted for 3 of the 6 quality losses.
1
Sources
50%
Confidence
Long-term
Relevance
9/25/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedSwift-27B's accuracy on AIME 2026 drops from 98.7% to 94.0%, a decrease of 4.6%, and the team acknowledges this is a bug in training where a critical math reasoning token was mistakenly penalized73% similarUnverifiedWhile maintaining nearly equivalent quality, the Swift version compresses both reasoning token consumption and latency to approximately 40% of the base version, Qwen3.8-Flash-Next.67% similarUnverifiedEarly testing on the Opus model showed an exact string recognition rate of 0 out of 15, whereas after the release of Fable 5 the same test yielded 13 out of 15.61% similarUnverifiedResearch shows that 30B-parameter models have a success rate 15–30 percentage points lower than GPT-4/Claude-3.5 on complex reasoning tasks.60% similarVerifiedThe range of SwiftUI's minimumScaleFactor parameter is 0 to 1, representing the minimum scale to which the text is allowed to shrink60% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/951067API
curl https://kongchang.com/api/v1/knowledge/claims/951067MCP
get_claim(id=951067)