Models

Each model runs one fixed recipe: the prompt and settings are versioned, and change only with a new version. Prices are per million tokens, as charged. Reasoning tokens count as output.

Model	Version	Input	Output	Most per page	ParseBench
Claude Opus 5.5 anthropic/claude-opus-5-5	2026-09-25	$4.40	$22.00	$0.86	85.71
Gemini 3 Flash (thinking high) google/gemini-3-flash	2026-09-28	$0.55	$3.30	$0.22	82.29
Gemini 3.8 Flash (thinking low) google/gemini-3.8-flash-low	2026-09-28	$0.83	$4.13	$0.14	73.68
GPT-5.6 Terra (reasoning none) openai/gpt-5.6-terra	2026-09-25	$2.20	$13.20	$0.48	78.11
Infinity-Parser2-Flash infly/infinity-parser2-flash	2026-09-29	$0.19	$0.96	$0.04	65.12

Most per page is the highest a page can ever cost. A request needs that much available per page to start, and is then charged only for what its pages used; the rest is released right away.

Up to 50 pages per request. Prices as of version 2026-09-29.