Unbiased AI's blended multimodal LLM — it runs several models in parallel on the same request and returns the synthesized best answer, with a 1M-token context.
curl https://api.venice.ai/api/v1/chat/completions \
-H "Authorization: Bearer $VENICE_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "pareto",
"messages": [{ "role": "user", "content": "Explain quantum tunneling simply." }]
}'Overview
Pareto is Unbiased AI's blended multimodal LLM. Instead of routing your request to a single model, it runs several in parallel and returns a synthesized best answer, without switching models mid-conversation. It accepts text and image input, supports tool calling and web search, and offers a 1M-token context for research, coding, and agentic workflows.
On Venice, Pareto runs under the anonymized privacy tier — your prompts are not stored, profiled, or fed into a personal history, which matters when you're piping source code or research material through a composite model. You pay per token ($1 in · $4 out per 1M, with cached input at $0.04) instead of a subscription, and the same blended behavior is available through the Venice app or the OpenAI-compatible API. Note that this tier is anonymized rather than end-to-end encrypted or TEE-isolated.
Agent quickstart
The API is OpenAI-compatible: change the base URL and the model id and existing client code works unchanged.
curl https://api.venice.ai/api/v1/chat/completions \
-H "Authorization: Bearer $VENICE_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "pareto",
"stream": true,
"messages": [{ "role": "user", "content": "Draft the release note." }]
}'curl https://api.venice.ai/api/v1/chat/completions \
-H "Authorization: Bearer $VENICE_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "pareto",
"messages": [{ "role": "user", "content": "Find the rate limits." }],
"tools": [{
"type": "function",
"function": {
"name": "search_docs",
"description": "Search the API documentation.",
"parameters": {
"type": "object",
"properties": { "query": { "type": "string" } },
"required": ["query"]
}
}
}],
"tool_choice": "auto"
}'import os
from openai import OpenAI
client = OpenAI(
api_key=os.environ["VENICE_API_KEY"],
base_url="https://api.venice.ai/api/v1",
)
resp = client.chat.completions.create(
model="pareto",
messages=[{"role": "user", "content": "Build without permission."}],
)
print(resp.choices[0].message.content)Specifications
Assessment
Use cases
Prompting
Keep your system prompt and conversation history stable across turns — reused input bills at the $0.04/1M cached rate instead of $1.
Attach screenshots, charts, or UI captures as image input rather than describing them; Pareto handles vision natively.
Define your tool schemas up front and let function calling drive multi-step agent loops instead of narrating steps in prose.
For anything time-sensitive, explicitly ask Pareto to use web search rather than relying on training data.
Dump entire codebases or document sets in one prompt — the 1M context is there to be used, and cached history keeps re-reads cheap.
Ask for the outcome, not for a specific underlying model's style — the blend picks the strongest response, so style directives can fight the synthesis.
Alternatives
| Model | Best for | Context | Open weights | Price (Venice) |
|---|---|---|---|---|
| Pareto | Blended multi-model answers | 1M tokens | No | $1 in · $4 out / 1M |
| Gemini 3.8 Flash | Fast general-purpose chat | 1M tokens | No | $0.94 in · $4.69 out / 1M |
| Muse Spark Open | Open-weight long-context work | 1M tokens | Yes | $1.56 in · $5.31 out / 1M |
| Qwen 3.7 Plus | Budget 1M-context tasks | 1M tokens | No | $0.50 in · $2 out / 1M |
Pareto is the right pick when you want one endpoint that quietly blends several models and returns the strongest answer — especially for research and agentic pipelines that lean on cached 1M-token context. If the lowest token price or open weights matter more than blended quality, Qwen 3.7 Plus or Muse Spark Open are the better fit.
Pricing
Billed per token on Venice: $1 per 1M input tokens and $4 per 1M output tokens.
New Venice accounts include a free daily allowance and 500 welcome credits — no credit card required.
Getting a key
Create a key in API settings. Nothing else is required to start.
Export it as VENICE_API_KEY so the snippets above run unedited.
Point an existing OpenAI client at https://api.venice.ai/api/v1. The scheme is part of the value: an OpenAI client given a bare host does not resolve it.
Pass pareto as the model and send the request.
FAQ
Pareto is a blended multimodal LLM from Unbiased AI, built by the remote-first team Circuit & Chisel. Rather than routing a request to one model, it runs several models in parallel against the same request and returns a synthesized best answer, without switching models mid-conversation. It accepts text and image input and targets research, coding, and agentic workloads.
On Venice, Pareto is billed per token: $1 per 1M input tokens, $4 per 1M output tokens, and cached input at just $0.04 per 1M tokens. There is no subscription requirement — you pay for the tokens you use.
Pareto is neither free to self-host nor open source — it is a closed, proprietary model with no published weights, so it cannot be downloaded or fine-tuned. It is available per-token on Venice, where new accounts get free welcome credits to try it.
Both offer a 1M-token context at nearly identical Venice pricing. Pareto's edge is its blended architecture — several models run in parallel and the best answer is synthesized — plus cached input at $0.04/1M. Gemini 3.8 Flash is the single-model pick when raw speed on general chat matters most.
Yes. Pareto accepts images alongside text, so you can send screenshots, charts, or diagrams for visual analysis. It outputs text only — it does not generate images, audio, or video.
Yes. Pareto supports function calling for building agents and tool-driven pipelines, and it can use web search for up-to-date information. Third-party listings also report structured-output support, though the dedicated reasoning mode is not among its capabilities.
Venice runs Pareto under its anonymized privacy tier: prompts are not stored, profiled, or tied to a personal history. This is anonymized handling rather than end-to-end encryption or TEE isolation, but it means your research material and code don't accumulate into a profiled account.
No dedicated reasoning mode is listed for Pareto. Its quality comes from the blend — multiple models answering in parallel with the best response synthesized — rather than from extended chain-of-thought. If explicit reasoning traces are essential, pair it with a reasoning-focused model on Venice.
One key, free to start, no credit card.