Solar Pro 3
Solar Pro 3 is Upstage’s Mixture-of-Experts large language model with 102B total parameters (12B active), a 128K-token context window, and strong extended reasoning and tool-use capabilities.
What is Solar Pro 3?
Solar Pro 3 is a proprietary Mixture-of-Experts language model from Upstage optimized for efficient, high-quality text generation and reasoning. It is used for complex multi-step reasoning, agentic workflows, and long-context tasks such as document analysis and large-codebase assistance. It also serves enterprise applications that need reliable tool use, structured outputs, and multilingual support focused on Korean with additional English and Japanese coverage. Solar Pro 3 follows earlier Solar-series models such as Solar Pro 2, offering increased parameter scale and improved reasoning performance within the same general model family.
Providers
Different companies host the same model. OpenRouter routes your request to one of them based on the routing mode you pick — Balanced (price + speed), Nitro (fastest), or Exacto (highest tool-calling accuracy).
| Provider | Input | Output | Cache read /M | Latency | Throughput | Uptime |
|---|---|---|---|---|---|---|
| Upstage | ~$0.25 | ~$0.75 | — | ~250ms | ~40 tps | ~99.9% |
| OpenRouter | ~$0.30 | ~$0.90 | — | ~320ms | ~35 tps | ~99.9% |
| Together AI | ~$0.28 | ~$0.85 | — | ~280ms | ~45 tps | ~99.9% |
| Fireworks AI | ~$0.26 | ~$0.80 | — | ~260ms | ~50 tps | ~99.95% |
Try this model
Test Solar Pro 3 right here — free to start.
Suggestions for your first prompt
Code snippet
Call the model through the OpenAI-compatible API.
from openai import OpenAI
client = OpenAI(
api_key="YOUR_API_KEY",
base_url="https://inference.example.com/v1"
)
response = client.chat.completions.create(
model="upstage/solar-pro-3",
messages=[
{
"role": "user",
"content": "Describe this image in one sentence."
}
],
)
print(response.to_json())
{
"model": "upstage/solar-pro-3",
"messages": [
{
"role": "user",
"content": "Describe this image in one sentence."
}
]
}
5 Core Capabilities
-
Text Generation
Generates and edits high-quality text responses across domains, suitable for content creation, SEO workflows, and structured writing tasks.
-
Long Context Handling
Processes and reasons over long inputs with a context window up to 128K tokens, supporting document-heavy and retrieval-oriented applications.
-
Tool Use
Supports tool use and function calling, enabling agentic workflows that interact with external systems and APIs programmatically.
-
Structured Outputs
Produces well-structured JSON and schema-conformant outputs, useful for automation pipelines and programmatic integration with downstream systems.
-
Multilingual Support
Handles multiple languages with strong performance in Korean and solid English and Japanese support for multilingual applications.
6 Most Valuable Use Cases
- Code Generation Assistance
- Enterprise Document Search
- Contract Review Support
- Invoice Extraction Automation
- Customer Support Triage
- Business Process Workflows
Why Build on LLM.API?
One unified API. Every major model. Built-in reliability, cost control, and observability.
-
Unified AI Routing
Dynamically route each request to the optimal model across providers using latency, cost, and quality signals—without changing your integration or redeploying code.
One endpoint, every model. -
Smart Cost Controls
Define budgets, price ceilings, and routing rules so LLM.API automatically picks the cheapest viable model while preserving output quality and performance SLAs.
Optimize spend by default. -
Automatic Fallbacks
Guard against provider outages and rate limits with configurable failover logic that instantly retries on backup models, maintaining uptime without custom error-handling glue.
Resiliency built in. -
Deep Observability
Get per-request traces, latency and cost breakdowns, and structured logs across all providers from a single dashboard and API, ready for alerting and analytics.
One pane of glass. -
Task-Aware Orchestration
Describe tasks at a higher level—chat, extraction, tools—and let LLM.API select prompts, parameters, and models, so you ship features instead of tuning configs.
Think tasks, not prompts. -
High-Throughput Batch
Send massive batches of jobs through a single API call with concurrency controls, retries, and progress tracking, ideal for backfills, evaluations, and bulk processing.
Scale jobs, not code.
When to Use — When NOT to Use
Use it if...
- You need a strong general-purpose model for chatbots and virtual assistants.
- You need solid coding assistance, including code completion, debugging, and explanation tasks.
- Your use case involves multilingual text understanding and generation across many major languages.
- Your use case involves drafting, rewriting, and polishing emails, reports, and marketing copy.
- You need a capable model for question answering over moderately long documents or webpages.
- You need a balance between quality and cost for everyday enterprise productivity workflows.
Avoid if...
- You need cutting-edge performance on the hardest reasoning or math competition benchmarks.
- Your workload requires guaranteed support for extremely long contexts, like hundreds of thousands tokens.
- You need tightly integrated image or multimodal capabilities beyond basic text-only interactions.
- You need deterministic, fully reproducible outputs with strict token-by-token compatibility guarantees.
- Your workload requires highly specialized domain models, like medical diagnosis or legal argumentation.
- You need robust offline deployment on highly constrained edge devices with minimal hardware resources.
Frequently Asked Questions
-
What is Solar Pro 3?
Solar Pro 3 is a large language model by Upstage optimized for high-quality reasoning, coding, and general-purpose chat via the LLM.API gateway.
-
What is Solar Pro 3 best used for?
Solar Pro 3 is best for complex reasoning, code generation and debugging, multi-step tool use, and production-grade chatbots needing strong instruction following.
-
What is the context window of Solar Pro 3?
Solar Pro 3 supports a long context window suitable for large documents and multi-step conversations; check the LLM.API model card for the exact token limit.
-
How fast is Solar Pro 3 in terms of latency and throughput?
Typical end-to-end latency is on the order of seconds for short prompts, with streaming responses and scalable throughput handled by LLM.API infrastructure.
-
Which modalities does Solar Pro 3 support?
Solar Pro 3 is a text-only model that accepts text prompts and returns text completions or chat responses.
-
How do I call Solar Pro 3 through LLM.API?
You can select the upstage/solar-pro-3 model name in the LLM.API completion or chat endpoint, passing your prompt and any temperature or max_tokens parameters.
-
How is Solar Pro 3 priced on LLM.API?
Solar Pro 3 uses pay-as-you-go, per-token billing; see the LLM.API pricing page for current input and output token rates.
-
How does Solar Pro 3 compare to similar models?
Solar Pro 3 is positioned as a high-quality, cost-efficient general model competitive with other top-tier reasoning and coding LLMs in its price bracket.
-
What are the main limitations of Solar Pro 3?
Solar Pro 3 can hallucinate, lacks real-time knowledge or browsing, and may underperform on highly specialized domain tasks without careful prompting or grounding.
-
Can I fine-tune or customize Solar Pro 3 via LLM.API?
Fine-tuning support depends on LLM.API capabilities at the time; check the model page for whether custom fine-tunes or adapters are available for Solar Pro 3.
COMPARE
Competitive Models
-
Nova 2 Lite
Nova 2 Lite is an Amazon foundational language model variant designed to provide efficient, general-purpose AI capabilities with reduced computational footprint. It is intended for everyday workloads where cost-effectiveness and responsiveness are prioritized over maximum scale.
-
Kimi K2.6
Kimi K2.6 is MoonshotAI’s open-source, 1-trillion-parameter Mixture-of-Experts multimodal model optimized for long-horizon coding, agentic tool use, and image/video understanding. It is notable for its large ~262K-token context window and strong performance on complex software engineering and tool-using benchmarks.
-
DeepSeek V4 Flash (free)
DeepSeek V4 Flash (free) is an open-source, efficiency-optimized Mixture-of-Experts language model from DeepSeek, offering a 1M-token context window with only 13B parameters activated per token out of 284B total. It is designed to deliver fast, cost-effective long-context reasoning, coding, and agentic workflows.
Get one key to every model
Swap your API key. Keep your code.