Pricing

Pay for results, not attempts.

Three ways to buy, one prepaid balance, one API key. Top up what you want to spend — when it runs out, requests stop. There is no invoice at the end of the month and no way to overspend.

Tasks

Document extraction

$0.010per page

Send a PDF, DOCX or plain text with a JSON Schema. Get back data matching that schema. A document we cannot read bills nothing.

$0.006 per page on the batch tier.

Agents

Issue to pull request

$5.00per pull request

Point it at a repository and an issue. It opens a pull request. Billed only when a PR opens with a real diff — a failed run costs nothing.

Inference

An OpenAI-compatible endpoint. Change your base URL, keep your code. Billed per 1M tokens on the tokens the model actually reports.

Model	Context	Input / 1M	Output / 1M
gpt oss 120b openai	131K	$0.036	$0.167
DeepSeek V4 Flash 0731 deepseek-ai	164K	$0.059	$0.176
DeepSeek V4 Flash deepseek-ai	164K	$0.088	$0.176
Llama 3.3 70B Instruct meta-llama	131K	$0.132	$0.392
Kimi K2.7 Code moonshotai	262K	$0.666	$3.33
Kimi K2.6 moonshotai	262K	$0.735	$3.43
DeepSeek V4 Pro deepseek-ai	164K	$1.27	$2.55

Flat rate: the same price at every context length, with no surcharge for long prompts and no time-of-day pricing.
Inference is billed on actual input and output tokens reported by the model.
Tasks and Agents are billed per unit — a page, a pull request — never per token.
A failed run bills nothing. You are charged for results, not attempts.

Live from the API · generated Sun, 27 Sep 2026 23:53:01 GMT