Solar Pro 3
Solar Pro 3 by Upstage is a Mixture-of-Experts LLM with a 128K context window and Korean-optimized multilingual support for reasoning and tool use.
Model Overview
Capabilities, design details, and architectural traits
Solar Pro 3 - Multilingual Mixture-of-Experts LLM
Solar Pro 3 is presented by Upstage as a Mixture-of-Experts language model designed for reasoning, tool use, and multilingual performance, with a strong focus on Korean.
Defining Characteristics
- Mixture-of-Experts architecture: Officially described as a MoE model, emphasizing computational efficiency.
- 128K context window: The model is documented with a 128K token context window, supporting long prompts and extended workflows.
- Korean-first multilingual support: Upstage positions Solar Pro 3 as optimized for Korean, with English and Japanese support also documented.
- Tool use and structured outputs: Official and provider documentation mention tool use, function calling, and structured output/schema-compliant generation.
Core Design
- Solar Pro 3 is framed as a reasoning-oriented enterprise model rather than a general chat model, with agent-like workflows called out in official partner documentation.
- The model is available through Upstage-related platforms and partner integrations, including Puter.js and OpenRouter listiter.js and OpenRouter listings.
Benchmark Performance
Independent evaluations · Artificial Analysis
Accuracy & Capability Details
Compare Models Side-by-Side
Evaluate specifications, pricing, and independent benchmark indices
| Model Details | |||
|---|---|---|---|
| General Info | |||
| Provider | Upstage | Anthropic | Anthropic |
| Release Date | April 6, 2026 | September 22, 2026 | September 28, 2026 |
| Knowledge Cutoff | - | - | Jun 2026 |
| Context & Limits | |||
| Context Window | 128K | 1M Best Context Window | 1M Best Context Window |
| Pricing (per 1M tokens) | |||
| Input Pricing | $0.15 Best Input Pricing | $4 | $2 |
| Output Pricing | $0.60 Best Output Pricing | $20 | $10 |
| Modalities | |||
| Inputs | text | textimagefile | textimagefile |
| Outputs | text | text | text |
| Benchmarks (0-100) | |||
| Intelligence Index | 7.8 | 57.6 Best Intelligence Index | 56.0 |
| Coding Index | 16.2 | - | - |
| Agentic Index | 1.4 | - | - |
Humanity's Last Exam
Extremely difficult logical reasoning and knowledge.
Long Context Reasoning
Logical reasoning over long context windows.
SciCode Benchmark
Scientific coding and mathematical modeling.
Independent evaluation data provided by Artificial Analysis. To view the latest benchmarks and full details, visit their official site.
Explore more from Upstage
Other models by Upstage
Top AI Models
Leading alternatives by intelligence score
