Is Solar Pro 4 available with Zero Data Retention on Elosia?

Solar Pro 4 runs on Elosia's standard backend; ZDR is available on many other models in the catalog.

How much does Solar Pro 4 cost on Elosia?

Solar Pro 4 is billed by usage at $0.03 per 1M input tokens and $0.12 per 1M output tokens, within your Elosia plan.

What is Solar Pro 4's context window?

Solar Pro 4 supports a context window of up to 524,288 tokens.

Is Solar Pro 4 hosted in the EU?

Solar Pro 4 is not available on Elosia's EU-hosted (EdenAI) backend; it runs on the default backend.

When should you choose Solar Pro 4?

Solar Pro 4 is a balanced model: a strong quality/cost trade-off for everyday production work.

Solar Pro 4

UpstageBalanced

ThinkingTool UseStructured Output

About this model

Solar Pro 4 is Upstage's cost-efficient large language model, featuring a 524K context window. It is built for long-horizon tasks and agentic workflows, with strong capabilities in office productivity, document-intensive...

Performance Tier

Balanced

Solar Pro 4 is a balanced model from Upstage : strong performance at a reasonable price.

Strong cost-performance ratio. Reliable for most professional use cases without premium pricing.

Pricing

This model is included in Elosia plans

Eco

Minimal cost. Ideal for very high volume or simple tasks.

Type	per 1M tokens
Input (prompt)	$0.030
Output (completion)	$0.120
Cache read	$0.0060

Capabilities

Context Length524K

Max Output Tokens131K

TokenizerOther

Inputtext

Outputtext

Release DateAugust 10, 2026

Benchmarks

General Intelligence

MMLU

Not reported

MMLU-Pro

86.3%

GPQA Diamond

89%

Mathematics

MATH-500

Not reported

AIME 2026

95.3%

Programming

HumanEval

Not reported

SWE-bench Verified

70.6%

LiveCodeBench

87.8%

Agentic

Terminal-Bench 2.1

57%

Where does Solar Pro 4 stand?

Compare its performance index against every other model.

View the performance leaderboard

Recommended Use Cases

CodingAnalysisTranslationGeneral Chat

Strengths

512K-token context with up to 128K of output in a single response
Tool calling supported on the API, for long-horizon agentic loops rather than single-turn answers
Competition mathematics, code generation and general reasoning are all part of the reported suite, so the scope is not code alone
English, Korean and Japanese supported, with a documented Korean and Japanese specialisation covering medical and legal domain suites

Limitations

Agentic execution, precisely what the model claims, is its softer axis next to its own reasoning results
Most of the published benchmark rows are self-evaluated in house rather than measured by a third party. Hard scope limits on top: text only, three languages, closed weights