Skip to content
Claude Fable 5.1The point release that mattered more than most whole numbers, and the first model that reliably says it is unsure instead of guessing well.1- GPT-5.6 SolGPT-5.6 SolOpenAI's July flagship is the model to beat on agentic coding, and the one you supervise least2
Claude Opus 5.5Fable-level work at a lower Opus price, and on agentic coding the new number to beat.3
ChatGPT Sol 5.6The best coding model anyone sells, and it gets there on roughly half the tokens and a third less money.4
Claude Sonnet 5.5Opus-class agentic coding at half the Opus price, and on Terminal-Bench the best score Anthropic has published.5
CursorA fork of the editor rather than a plugin for it, which is why it can do things a plugin cannot.6
Claude Haiku 5.5A small model that outworks last year's big ones on agent tasks, at ten cents per million tokens in.7
GPT-6.1 SolClose to GPT-6 Astra on agentic work for a fraction of its cost, a week after the Sol it replaced.8
Claude Haiku 4.5The small model that stopped being the one you settle for, and the right default for anything you run a thousand times.9
GitHub CopilotThe autocomplete that started the whole category, still the one that fits most naturally into a working day.10
