OpenAI · released
GPT-5.6 Sol
The broadest published scoreboard of any model right now.
Overview
The top tier of OpenAI's three-model July release. Sol is the one that still publishes the classic academic benchmarks, which makes it the easiest frontier model to compare against anything else. It also posted the highest recorded throughput of the group on Cerebras hardware.
Solthar DNA rating, out of 5
- reasoning: 5
- coding: 4.5
- writing: 4.5
- vision: 5
- speed: 4.5
- value: 3.5
- context: 5
- tools: 4.5
Verdict
The safest pick when you need one model to be competent at everything.