Unverified50% confidenceBenchmarkExact time
Previously, models such as Gemini 2.5 Pro and GPT-5 could only achieve 10%–20% on Frontier Math Tier 4.
1
Sources
50%
Confidence
Long-term
Relevance
10/4/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedOn difficult math problems, both SOL and TERRA achieved perfect scores of 10, and on the adaptive fine-tuning task all three GPT-5.6 models achieved perfect scores, whereas the previous GPT-5.5 and Opus 4.7 could only score two to three points on the same task.78% similarUnverifiedGPT-5.4 was reversed by a "trusted source" on 44.7% of questions, Grok-4.20 had a reversal rate as high as 87.5%, while Gemini-3.1-Pro was nearly unaffected at just 0.6%.76% similarUnverifiedEarly GPT-4 achieved a resolution rate of less than 2% on SWE-bench, while recent top models have surpassed 50%.73% similarUnverifiedGPT-4、Gemini Ultra等大型语言模型在竞赛数学上已能达到接近顶级学生的水平，但在研究级证明上仍频繁出错73% similarPartially VerifiedAll three GPT-5.6 models achieved a perfect score of 10 on the long-horizon agent fine-tuning task, while GPT-5.5 and Opus 4.7 could only score two or three points on that task.72% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/972625API
curl https://kongchang.com/api/v1/knowledge/claims/972625MCP
get_claim(id=972625)