Coverage (30d)
7vs1
This Week
3vs0
Evidence
2 articlesRelationships
0Timeline
GPT-5.2 Pro2026-04-29
GPT-5.5 Pro shows significant leap on Epoch benchmark; base GPT-5.5 surpasses GPT-5.4 Pro
GPT-5.2 Pro2026-04-23
Ethan Mollick shares early impressions of GPT-5.5 Pro, calling it the best model for hard problems.
GPT-5.2 Pro2026-04-19
OpenAI insider describes GPT-5.5 Pro as 'qualitative leap' and 'absolutely insane'
GPT-5.2 Pro2026-04-19
Reported to generate complex, high-quality SVG code from a single prompt
GPT-5.2 Pro2026-03-23
Reportedly solved an open problem in FrontierMath with human verification
GPT-5.4 Thinking2026-03-14
New model unveiled, reportedly matches human experts with 83% score on GDPVal benchmark
GPT-5.2 Pro2026-03-08
Identified as likely the best model in the world for catastrophic risk-relevant tasks including bioweapons research and cyber operations
Ecosystem
GPT-5.2 Pro
usesscientific discovery1 src
deploysChain-of-Thought Prompting1 src
deploysInstruction Tuning (FLAN)1 src
deploysReinforcement Learning from Human Feedback (RLHF)1 src
deploysTransformer Self-Attention1 src
GPT-5.4 Thinking
usesCoT controllability1 src
Benchmarks
swe bench pro
GPT-5.2 Pro55.6
GPT-5.4 Thinking—
swe bench verified
GPT-5.2 Pro80
GPT-5.4 Thinking—