FrontierObservatory

Visual comparison

The frontier, plotted.

Filter by time, company, price and score; highlight a model, a group, or a whole lab. Every chart reads from the same filter rail.

SWE-bench Verified over timerelease date × score
xAIAnthropicGoogle DeepMindOpenAIMeta AIMistral AIDeepSeekMoonshot AIAlibaba (Qwen)Zhipu AISoofi
Cost vs SWE-bench Verifiedinput $/MTok (log) × score — up and left is the frontier
xAIAnthropicGoogle DeepMindOpenAIMeta AIMistral AIDeepSeekMoonshot AIAlibaba (Qwen)Zhipu AISoofi

Head-to-head picks (4/5)

Head to headlaunch-time reported scores; missing bars = not reported