NOUMENOpen app
NOUMEN

upstage · Efficient

Solar Pro 3

Added Mar 3, 2026

Compare models

Choose up to 3 models to compare.1 of 3 chosen

Loading models…

Use in the app

Upstage's Solar Pro 3 is a Mixture-of-Experts (MoE) language model with 102B total parameters and 12B active parameters per forward pass, optimized for Korean with strong English and Japanese support.

Input
Text
Output
Text
Discovery tier
Efficient

Pricing

USD per 1M tokens
Input
$0.15
Output
$0.60
Cache read
$0.015

Reference prices as published by the model provider. API example

Specifications

Context window
131.1K
Max output
—

Side by side

Solar Pro 3 compared

Specifications, prices, speed and benchmark scores next to each other. Benchmarks below show every compared model too.

Compare models

Choose up to 3 models to compare.1 of 3 chosen

Loading models…

Solar Pro 3upstageThis modelQwen2.5 72BQwenRemove
ProviderupstageQwen
Discovery tierEfficientOpen weights
AddedMar 3, 2026Jul 3, 2025
Limits
Context window131.07K131.07K
Max output—8.19K
Average output—166 tokens
Pricing, USD per 1M tokens
Input$0.15Best$0.36
Output$0.60$0.40Best
Cache read$0.015—
Benchmark indexes
Intelligence Index7.8Best7.7
Coding Index16.2—
Math Index—14.0
Capabilities
ToolsNoYes
Reasoning effort——

Public evaluations

Benchmarks

Independent evaluation results, with how each score ranks among the models in the Noumen catalog.

  • Solar Pro 3
  • Qwen2.5 72B

Intelligence Index

  • Solar Pro 37.8
  • Qwen2.5 72B7.7

Solar Pro 3: Better than 12% of 326 scored models

3.8Catalog models by score55.1

Indexes

  • Coding Index

    Composite of the coding evaluations

    • Solar Pro 3: 16.2
    • Qwen2.5 72B: —

  • Math Index

    Composite of the math evaluations

    • Solar Pro 3: —
    • Qwen2.5 72B: 14.0

Agentic work

  • AutomationBench-AA

    Workflow automation, penalized for guardrail violations

    • Solar Pro 3: 0.5%
    • Qwen2.5 72B: —

  • AutomationBench-AA tasks completed

    Workflows finished end to end without a guardrail violation

    • Solar Pro 3: 0.0%
    • Qwen2.5 72B: —

  • AA-Briefcasev1.1

    Agentic knowledge work, rated head to head

    • Solar Pro 3: 99 Elo
    • Qwen2.5 72B: —

  • GDPval-AAv2.1

    Economically valuable occupational tasks, rated head to head

    • Solar Pro 3: 264 Elo
    • Qwen2.5 72B: —

Document reasoning

  • GDP.pdf

    Professional reasoning over PDF documents, all-pass rate

    • Solar Pro 3: 1.6%
    • Qwen2.5 72B: —

  • AA-LCRv1.1

    Reasoning across long documents

    • Solar Pro 3: 32.3%
    • Qwen2.5 72B: —

  • MLCR-AA

    Long-context reasoning over medical records

    • Solar Pro 3: 0.6%
    • Qwen2.5 72B: —

Reasoning and knowledge

  • Humanity’s Last Exam

    Expert-level questions across academic fields

    • Solar Pro 3: 10.3%Top
    • Qwen2.5 72B: 3.6%

  • GPQA Diamond

    Graduate-level science questions

    • Solar Pro 3: 72.4%Top
    • Qwen2.5 72B: 49.1%

  • MMLU-Pro

    Multi-subject knowledge and reasoning

    • Solar Pro 3: —
    • Qwen2.5 72B: 72.0%

  • IFBench

    Following precise output instructions

    • Solar Pro 3: 71.2%Top
    • Qwen2.5 72B: 36.9%

Coding

  • Terminal-Benchv4.0

    Practical software engineering in a terminal

    • Solar Pro 3: 0.0%
    • Qwen2.5 72B: —

  • SciCode

    Scientific computing in Python

    • Solar Pro 3: 25.5%
    • Qwen2.5 72B: —

  • LiveCodeBench

    Recent competitive programming problems

    • Solar Pro 3: —
    • Qwen2.5 72B: 27.6%

Math

  • AIME 2025

    Competition mathematics, 2025 problems

    • Solar Pro 3: —
    • Qwen2.5 72B: 14.0%

  • AIME

    Competition mathematics

    • Solar Pro 3: —
    • Qwen2.5 72B: 16.0%

  • MATH-500

    Competition math problems

    • Solar Pro 3: —
    • Qwen2.5 72B: 85.8%

Legacy evaluations

  • Terminal-Bench Hard

    Earlier terminal task suite

    • Solar Pro 3: 7.6%Top
    • Qwen2.5 72B: 4.5%

  • τ²-Bench Telecom

    Tool use in a customer-support simulation

    • Solar Pro 3: 86.3%Top
    • Qwen2.5 72B: 34.5%

  • AA-LCR (unversioned)

    Long-context reasoning, earlier grading

    • Solar Pro 3: 32.3%
    • Qwen2.5 72B: —

Scores from Artificial Analysis, updated Oct 26, 2026. Artificial Analysis

Developers

Use it through the API

Noumen’s API is OpenAI-compatible: any Chat Completions client works with a base URL and a key.

Read the quickstart
curl https://www.noumen.xyz/api/v1/chat/completions \
  -H "Authorization: Bearer $NOUMEN_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "upstage/solar-pro-3",
    "messages": [{ "role": "user", "content": "Explain this stack trace." }]
  }'
import os
from openai import OpenAI

client = OpenAI(base_url="https://www.noumen.xyz/api/v1", api_key=os.environ["NOUMEN_API_KEY"])
reply = client.chat.completions.create(
    model="upstage/solar-pro-3",
    messages=[{"role": "user", "content": "Explain this stack trace."}],
)
print(reply.choices[0].message.content)

upstage

Other upstage models, starting with the same family.