Unverified50% confidenceFactExact time
According to the Reddit post description, GPT-6.1 Sol uses 2 inference passes rather than three.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
10/6/2026
First Seen
Valid until: 1/4/2027
Sources
Related Entities
Related Claims
UnverifiedWhen pressed twice (a double gaslighting test), both Opus 5.5 and GPT-6 Sol held their ground without arbitrarily changing their answers, clearly explaining why they maintained their judgments.72% similarUnverifiedGPT-5.6 Sol achieves 92.2% on the BrowseComp browsing task benchmark.71% similarUnverifiedGPT-5.6 Sol引入了Max和Ultra两个新的推理等级70% similarVerifiedSOL's benchmark score is nearly double that of the previous generation GPT-5.5, representing a generational leap, but still trails Fable 5 overall.70% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/980684API
curl https://kongchang.com/api/v1/knowledge/claims/980684MCP
get_claim(id=980684)