LlamaIndex answers, written by Pareto

Q: Why are my requests going to /completions?

OpenAILike treats a model as a plain completion model unless you say otherwise. Pass is_chat_model=True so requests go to /v1/chat/completions.

Q: Why does the response say pareto-26.10?

That is the release of Pareto that answered. Keep sending pareto as the model name in your code.

LlamaIndex does the reading: it loads your documents, indexes them and hands the right passages to a model. Pareto can do the writing. Set Settings.llm to an OpenAILike object with the Unbiased base URL and the model name pareto, and your query engines and agents answer through it. Pareto selects frontier and open models for each request and blends their outputs into one answer.

Choose Personal · $10/month See the setup →

Tested live with LlamaIndex core 0.14.2550M tokens a week

ask.py · handbook

ask.py

from llama_index.core import (    Settings, SummaryIndex)from llama_index.core.query_engine \    import CitationQueryEnginefrom llama_index.llms.openai_like \    import OpenAILikeBASE = "https://api.unbiased.ai/v1"Settings.llm = OpenAILike(    model="pareto", api_base=BASE,    is_chat_model=True,    is_function_calling_model=True,    context_window=1000000)index = SummaryIndex.from_documents(    docs)qe = CitationQueryEngine.from_args(    index)print(qe.query(QUESTION))

TerminalPython

python ask.py

Loaded hours.md, parking.md, finance.md

Query When does the Harbor Street office open, and where do visitors park?

Visitors park in the blue garage on Pier Road [2]. The Harbor Street office opens at 8:30 am [1].

Source 1 hours.md

Source 2 parking.md

Source 3 finance.md

python ask.py

ask.pyllama-index-core 0.14.25 · pareto

A demonstration of LlamaIndex on Pareto, recreated from a real run. Not a screenshot.

Set it up with OpenAILike

Pareto speaks the OpenAI Chat Completions format, and LlamaIndex reaches endpoints like that through the llama-index-llms-openai-like package. Build one OpenAILike object, set it as Settings.llm, and every query engine, chat engine and agent that doesn’t name its own model uses Pareto.

Get an Unbiased API key
Choose a plan, then create a key in the Unbiased platform. New accounts are reviewed and approved by a person before the key works.
Install the OpenAILike integration
Run pip install llama-index-core llama-index-llms-openai-like. Use OpenAILike rather than llama_index.llms.openai.OpenAI: the OpenAI class expects OpenAI’s own model names, and pareto is not one of them.
Point it at Unbiased
Pass model="pareto", api_base="https://api.unbiased.ai/v1" (keep the /v1) and your key as api_key. Type pareto yourself; OpenAILike takes the name as given and does not look it up.
Set the three flags
is_chat_model=True sends requests to /v1/chat/completions instead of the older completions endpoint. is_function_calling_model=True lets FunctionAgent use native tool calls. context_window tells LlamaIndex how much text it can pack into one prompt; the default is far smaller than Pareto’s.
Choose an embedding model for vector search
Unbiased serves chat, not embeddings. A SummaryIndex needs none. A VectorStoreIndex needs Settings.embed_model set to another provider, such as a local Hugging Face model.

Tested on October 20, 2026LlamaIndex core 0.14.25 (llama-index-llms-openai-like 0.8.1) on Pareto (pareto). A 35-line script answered a question over three in-memory documents with a CitationQueryEngine, citing the two right sources, then a FunctionAgent called a multiply tool and returned the correct product. Run record.

Install

pip install llama-index-core llama-index-llms-openai-like

ask.py

from llama_index.core import Document, Settings, SummaryIndex
from llama_index.core.query_engine import CitationQueryEngine
from llama_index.llms.openai_like import OpenAILike

Settings.llm = OpenAILike(
    model="pareto",
    api_base="https://api.unbiased.ai/v1",
    api_key="your-unbiased-api-key",
    is_chat_model=True,
    is_function_calling_model=True,
    context_window=1000000,
)

# SummaryIndex needs no embeddings.
# A VectorStoreIndex also needs Settings.embed_model.
docs = [
    Document(text="The Harbor Street office opens at 8:30 am.",
             metadata={"file_name": "hours.md"}),
    Document(text="Visitors park in the blue garage on Pier Road.",
             metadata={"file_name": "parking.md"}),
    Document(text="Expense reports are due on the 5th.",
             metadata={"file_name": "finance.md"}),
]
index = SummaryIndex.from_documents(docs)
query_engine = CitationQueryEngine.from_args(index)

response = query_engine.query(
    "When does the office open, and where do visitors park?")
print(response)
for i, node in enumerate(response.source_nodes, 1):
    print(i, node.metadata.get("file_name"))

In real code, read the key from an environment variable rather than writing it into the file.

Your indexes stay. The writer changes.

One object in Settings changes. Your loaders, indexes and query engines carry on as written.

Readers, node parsers and indexesUnchanged
Query engines, citations and FunctionAgent toolsUnchanged
Embeddings for vector searchFrom another provider
Model stringpareto
Where requests goapi.unbiased.ai/v1
What you payFrom $10 a month

One model string, several models behind it

Pareto is Unbiased’s blended AI model. On each request it selects frontier and open models and blends their outputs into one answer, so LlamaIndex talks to one model string, pareto, and never has to pick.

“When the frontier improves, Pareto improves; when frontier prices fall, you should benefit.” From the pricing page.

How a Pareto answer gets made →

An illustration, not a wiring diagram. The blend includes frontier and open models, and its composition can change.

What an agent task costs on Pareto

Agents make many model calls per task, so the price of a finished task matters more than the price of a token. The model card publishes both for Pareto on two public agent benchmarks.

Terminal-Bench 4.0Tasks an agent completes in a terminal: $0.48/task
DeepSWE v1.1Changes an agent makes in a real codebase: $0.24/task

Pareto 26.10, preliminary October 1, 2026 runs that may change; denominators and the cost method still require confirmation. Cost is the reported mean USD per task at API rates, not an estimate for your workload. Model card and method.

Pay for LlamaIndex by the month

One API key works across LlamaIndex and any other tool you point at Pareto. Pick a monthly allowance, or pay per token on Team.

Personal

$10/month

50M tokens a week
20 requests per minute
For one person

Choose Personal

Personal Max

$100/month

600M tokens a week
60 requests per minute
For one person

Choose Personal Max

Team

Pay as you go

Prepaid credits at published rates
300 requests per minute
No weekly token cap

Get started with Team

Subscriptions are for one person and are not unlimited. New API accounts are reviewed by a person before access turns on, and commercial or production use needs a separate written agreement. Rates and limits are on Pricing.

LlamaIndex on Pareto, questions answered

Does LlamaIndex work with Pareto?

Yes, in our test. We ran LlamaIndex core 0.14.25 with llama-index-llms-openai-like 0.8.1 against Pareto on October 20, 2026: a CitationQueryEngine answered a question over three documents and cited the two that held the answer, and a FunctionAgent called a tool and returned the right result. Format compatibility is not a promise that every LlamaIndex feature works, so tell us if one doesn’t.

Can I build a VectorStoreIndex on Pareto?

Yes, with embeddings from somewhere else. Unbiased does not serve an embeddings endpoint, so set Settings.embed_model to another provider, such as a local Hugging Face model or a separate embeddings API. Pareto then writes the answers from whatever the retriever finds. A SummaryIndex needs no embeddings at all.

Why are my requests going to /completions?

OpenAILike treats a model as a plain completion model unless you say otherwise. Pass is_chat_model=True so requests go to /v1/chat/completions.

Why does the response say pareto-26.10?

That is the release of Pareto that answered. Keep sending pareto as the model name in your code.

Can I ship a LlamaIndex app on this plan?

Personal and Personal Max subscriptions are for one person. Commercial or production use needs a separate written agreement with Unbiased. If you are building agents or apps for others, see the page for agent builders.

Is Pareto made by the LlamaIndex team?

No. Pareto is Unbiased’s model. LlamaIndex is made by LlamaIndex, and Unbiased is not affiliated with or endorsed by them. You keep LlamaIndex and point it at the Unbiased API.

What happens when I hit the weekly allowance?

Requests over your plan’s limits return an HTTP 429 error. Personal includes 50M tokens a week and 20 requests per minute. Personal Max raises that to 600M a week and 60 requests per minute for $100/month. Team uses prepaid pay-as-you-go credits with no weekly token cap.

Do I choose between models?

No. You send every request to one model string, pareto. Pareto selects frontier and open models for each request and blends their outputs into one answer.

Is my code kept?

Personal and Personal Max are not our zero data retention tier: prompts and responses may be kept up to 30 days, as section 5(c) of the Terms explains. For work that must not be kept, ask about the Zero Data Retention API on the Private AI page.

Pareto works in other tools too

The same key and the same model string go into each of these. See them all.

Product names belong to their owners. Except where a guide says otherwise, Unbiased is not affiliated with or endorsed by the makers of the tools on this page.

Ask your documents something hard

One OpenAILike object, one key and the same indexes. Pareto starts at $10 a month for 50M tokens a week.