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.
Added Mar 3, 2026
Context Window
128.0K
Max Output
128.0K
Avg output tokens (7d)
24 tokens
Input Price (Auto)
$0.15/1M
Output Price (Auto)
$0.60/1M
Cache Read (Auto)
$0.015/1M
Benchmarks
Benchmarks
Performance metrics and benchmarks
Sourced from Artificial Analysis.
Intelligence Index
9.4
Coding Index
16.2
Agentic Index
1.4
Reasoning
GPQA Diamond
Graduate-level scientific reasoning
72.4%
Better than 52% of models compared
HLE
Humanity's Last Exam
10.3%
Better than 55% of models compared
IFBench
Instruction-following benchmark
71.2%
Better than 87% of models compared
T²-Bench Telecom
Conversational AI agents in dual-control scenarios
86.3%
Better than 79% of models compared
AA-LCR
Long context reasoning evaluation
32.3%
Better than 34% of models compared
GDPval-AA
Economically valuable tasks
0.0%
CritPt
Research-level physics reasoning
0.0%
Coding
SciCode
Python programming for scientific computing
25.5%
Better than 3% of models compared
Terminal-Bench Hard
Agentic coding and terminal use
7.6%
Better than 40% of models compared
Knowledge
AA-Omniscience Accuracy
Proportion of correctly answered questions
18.5%
AA-Omniscience Hallucination Rate
Rate of incorrect answers among non-correct responses
88.2%
Last updated Sep 7, 2026
Artificial AnalysisProviders
Auto routing is available for this model. Explicit provider selection is not available.
Loading provider options…
Related text models
Compare Solar Pro 3 with similar models from the same provider or model family.
Solar Pro 4
upstage/solar-pro4Upstage's Solar Pro 4 is a long-context language model for agentic workflows, office productivity, document-intensive work, and coding. This variant keeps reasoning disabled for faster direct responses.
Solar Pro 4 Thinking
upstage/solar-pro4:thinkingUpstage's Solar Pro 4 with reasoning enabled for harder agentic workflows, document-intensive work, coding, planning, and analysis across a 524K-token context window.
Synth 2.5 Pro Preview
synth-2.5-proSynth 2.5 Pro Preview is a low-cost text model designed for role-play, character dialogue, and collaborative storytelling.
GPT 6 Astra Pro
openai/gpt-6-astra-proGPT 6 Astra in Pro reasoning mode. Uses additional model work for difficult tasks, with higher latency and token usage at the same per-token rates. Reasoning effort remains independently configurable.
DeepSeek V4 Pro 0813
deepseek/deepseek-v4-pro-0813DeepSeek V4 Pro 0813 is the general-availability release of DeepSeek V4 Pro, built for coding, tool use, cybersecurity, automation, and long-horizon agent workflows. DeepSeek reports strong gains over the preview and scores above Opus 4.8 on Terminal Bench 2.1, Cybergym, DeepSWE, and AutomationBench. It supports a 1M-token context window.
DeepSeek V4 Pro 0813 Thinking
deepseek/deepseek-v4-pro-0813:thinkingDeepSeek V4 Pro 0813 Thinking enables reasoning by default on the general-availability release of DeepSeek V4 Pro, built for coding, tool use, cybersecurity, automation, and long-horizon agent workflows. DeepSeek reports strong gains over the preview and scores above Opus 4.8 on Terminal Bench 2.1, Cybergym, DeepSWE, and AutomationBench. It supports a 1M-token context window.