Unverified50% confidenceBenchmarkExact time
Even in the Indian context, 5% to 12% of responses from all models used Western thousands-separator formatting (Western comma drift) in intermediate steps, with Gemini 3.7 Flash the lowest (5%) and Claude Sonnet 4.6 the highest (12%).
1
Sources
50%
Confidence
Short-term (~14 days)
Relevance
10/11/2026
First Seen
Valid until: 10/25/2026
Sources
Related Entities
Related Claims
UnverifiedOn the Frontier Code programming benchmark, the old model scored 34.4%, while Gemini 3.7 Flash improved to 43.6%.62% similarUnverifiedOn the Frontier Code benchmark, GPT-6 Luna scored 56.1%, Claude Sonnet 5.0 scored 56.2%, and Kimi K3 scored 58.2%.61% similarVerifiedClaude Sonnet 5.5 achieved 70.6% on the Terminal benchmark, compared to only 10.3% for the previous generation.60% similarUnverifiedIn the easy-to-verify category, seven models scored a perfect 40/40 across all six bait types, and Claude Sonnet 5.5, Gemini 3.7 Flash, and GPT-6.1 Sol answered every question correctly across all six tasks in both the easy and hard categories.60% similarUnverifiedSonnet 5.5's reasoning tiers are divided into low/medium/high/x-high/Max, with token usage increasing only 5% to 8% from low to x-high, but jumping approximately 1500% (15×) from x-high to Max.59% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/996748API
curl https://kongchang.com/api/v1/knowledge/claims/996748MCP
get_claim(id=996748)