Roar AI
DocsRequest access
All models
Upstage · Chat

Solar Pro 4

Upstage’s model that says when it cannot verify something, rather than filling the gap.

Input$0.117per 1M tokens
Output$0.468per 1M tokens
BillingOne monthly invoiceevery model on one key
Request early accessSee every model’s rate

About Solar Pro 4

Solar Pro 4 is the commercial flagship of Upstage, the Korean AI company, rebuilt for agent work. Its defining trait is restraint: Upstage designed it to say it cannot verify a claim instead of inventing one, and to answer from the evidence in long, multi-document inputs. That roughly tripled its Artificial Analysis Intelligence Index score over Solar Pro 3, to 42.

Upstage officially supports English, Korean and Japanese, and reports 89.0% on GPQA Diamond. Some of its gains come from declining to answer rather than knowing more — which is exactly what you want when the documents are contracts and reports.

Where it’s strong

Questions over contracts, reports and other long documents.

Banking and customer-service agents that must not guess.

Korean and Japanese business documents.

Where to be careful

Slow on agent tasks: about 8.6 minutes per task in one independent test.

Independent testing found a usable context of about 384K tokens, short of the 512K Upstage states.

Well below the frontier models in absolute terms.

Price

Input$0.117 per 1M tokens
Output$0.468 per 1M tokens
Cached input$0.0234 per 1M tokens

For scale: 1,000 requests of about 3,000 tokens in and 1,000 out cost $0.82 — a long prompt and a page of answer each. Every rate is published and stays the same from one request to the next. See every model.

Specs

Made byUpstage
ReleasedAugust 2026
TypeChat
Context window524K tokens
Longest answer131K tokens
ReadsText
Thinks before answeringYes, adjustable
WeightsClosed
Model idsolar-pro4

Call it

Use the OpenAI SDK you already have — point it at Roar AI and name the model. Read the docs.

from openai import OpenAI

client = OpenAI(base_url="https://api.roar-ai.com/v1", api_key="roar_live_…")

reply = client.chat.completions.create(
    model="solar-pro4",
    messages=[{"role": "user", "content": "Summarise this contract in plain English."}],
)

Questions

How much does Solar Pro 4 cost?

Input is $0.117 per 1M tokens; output is $0.468 per 1M tokens. For scale, 1,000 requests of about 3,000 tokens in and 1,000 out cost $0.82. The rate is published and does not change from one request to the next.

Can I use Solar Pro 4 alongside other models?

Yes. Every model on Roar AI is on the same API key and the same monthly invoice, so switching from Solar Pro 4 to another model is a change to one word in your code.

Is Solar Pro 4 open source?

No. Upstage does not publish the weights, so it is only available through an API like this one.