Pareto is a multimodal composite model from Unbiased. Rather than being a single set of weights, it runs several models against the same request and returns the best answer, without switching mid-conversation — which is what lets prompt caching keep working across turns.
Unbiased positions it for research, coding and agentic workflows, with frontier-level quality across general-purpose tasks at a fraction of frontier cost. Pareto and the Unbiased platform are built by Circuit & Chisel, a remote-first team in the US and Canada.
Composite routing without losing the thread
Several models answer the same request and the best response is returned. Because the conversation does not switch models mid-flight, prompt caching survives.
A 262K context window
Long enough for whole repositories, large document sets, or extended agent transcripts, with up to 131,072 tokens back.
Image input
Pareto accepts images alongside text and returns text, so screenshots, diagrams and scanned pages can go straight into the prompt.
Agentic plumbing included
Tool calling, structured (JSON schema) output and streaming are all supported, which is what agent loops need to plan, call tools and iterate.
Coding and code review
Generate, debug and review code across languages, reading whole files or repositories in a single prompt.
Agentic and multi-step workflows
Reliable tool calling and structured output suit agent loops that plan, call tools and iterate.
Research over long sources
Summarise and extract from papers, contracts and reports without aggressive chunking.
Multimodal analysis
Read charts, screenshots and scanned documents alongside the surrounding text.
Sources: https://unbiased.ai/ · https://openrouter.ai/unbiased/pareto · https://circuitandchisel.com/
Measured on the AI/ML API gateway across live traffic. Figures are medians unless stated. Last updated 2026-10-12.
import requests r = requests.post( "https://api.aimlapi.com/v1/chat/completions", headers={"Authorization": "Bearer " + AIMLAPI_KEY}, json={ "model": "unbiased/pareto", "messages": [ { "role": "user", "content": "Hello!" } ] }, ) print(r.json())
const r = await fetch("https://api.aimlapi.com/v1/chat/completions", { method: "POST", headers: { Authorization: `Bearer ${process.env.AIMLAPI_KEY}`, "Content-Type": "application/json", }, body: JSON.stringify({ "model": "unbiased/pareto", "messages": [ { "role": "user", "content": "Hello!" } ] }), }); console.log(await r.json());
curl -X POST https://api.aimlapi.com/v1/chat/completions \ -H "Authorization: Bearer $AIMLAPI_KEY" \ -H "Content-Type: application/json" \ -d '{"model":"unbiased/pareto","messages":[{"role":"user","content":"Hello!"}]}'
OpenAI-compatible — swap the base URL and it works with your existing SDK.
| Type | Price |
|---|---|
| Input | |
| Output | |
| Cached input |
Pareto has a 262,144 token context window and can return up to 131,072 tokens.
Pareto takes image and text as input and returns text. Screenshots, diagrams and scanned pages can go into the prompt alongside the text.
Pareto is priced at input $3.4385 / 1M tokens, output $10.3155 / 1M tokens, cached input $0.34385 / 1M tokens.
Pareto is not a single set of weights. It runs several models against the same request and returns the best answer. Because it does not switch models mid-conversation, prompt caching keeps working across turns.
Yes. Pareto supports function calling, structured (JSON schema) output, streaming and vision — the set an agent loop needs to plan, call tools and iterate.
Pareto was built by Unbiased. Unbiased and Pareto are developed and maintained by Circuit & Chisel, a remote-first team across the US and Canada. The model was released on September 17, 2026.
Yes, Pareto can stream responses as they are generated.
Pareto is served on the OpenAI-compatible chat completions endpoint. Point any OpenAI SDK at https://api.aimlapi.com/v1 and set the model to unbiased/pareto. The short alias pareto also resolves to it.
