Solar Open 2: Korea's Sovereign Foundation Model, Built for Agentic Use
Upstage Studio — Build and deploy your agent.<br>Learn more →<br>Upstage Studio — Build and deploy your agent.<br>Learn more →
Solutions
Resources
Company
Pricing
Try now<br>Try demo<br>Contact us
Today we're releasing Solar Open 2, our open-weight foundation model optimized for agentic use.<br>Solar Open 2 is designed and trained to serve as the base model for agents in real work environments. It's trained on a wide range of agent scenarios, including tool calling, coding, and office work, and delivers performance that competes with leading global models across agent, Korean, knowledge, and coding benchmarks.<br>The model adopts an MoE architecture that activates only 15B of its 250B total parameters per token, improving capability and inference efficiency at the same time. With quantization, it runs on two NVIDIA H200 GPUs, which makes it practical for enterprises to deploy and operate on their own infrastructure. Built for agent tasks that run across many steps, it supports a context window of up to 1M tokens and comes with official support for Korean, English, and Japanese.<br>The model weights are released on Hugging Face under a commercially usable license. Alongside them, we publish the Solar Open 2 Technical Report, covering the full architecture, training methodology, key experiments, and evaluation conditions.<br>Key highlights<br>Designed and trained for agentic use, ready for real work environments<br>Competes with leading global models across agent, Korean, knowledge, and coding benchmarks<br>250B total / 15B active MoE architecture for efficient inference<br>1M-token context window<br>Runs on two H200 GPUs with quantization<br>Officially supports Korean, English, and Japanese<br>Released as open weights with commercial use permitted
From chat to agents<br>The role of LLMs is changing.<br>Until now, LLMs mostly generated a single answer to a user's question. They are now expanding into agents that search for information, read across documents, write code, and call external tools to carry real work through to completion.<br>The work agents handle is different in kind from chat. Instead of answering a short question, an agent repeats dozens of reasoning steps and tool calls. It has to work with information scattered across files and documents, remember what it has already done, and decide its next action accordingly. A mistake at one step can fail the entire task downstream.<br>Agentic use demands three capabilities.<br>First, long-horizon task execution : carrying multi-step work through to the end.<br>Second, long-context processing : holding long documents and work history in view.<br>Third, following instructions precisely and calling the right tools reliably.<br>These capabilities are hard to learn from ordinary question-answering data. The model needs to experience agent scenarios during training: search, tool calling, coding, document work. And the scenarios can't just look plausible. They have to reflect real work environments, and we have to be able to verify that the model actually completed the task.<br>Cost matters too. An agent doesn't stop at a single model call; it cycles through planning, execution, verification, and revision, so it typically uses far more tokens than chat. For an enterprise to deploy the model on its own infrastructure, the model can't be too large or too slow.<br>We built Solar Open 2 to be more than a model with high benchmark scores: a model with the performance and cost structure to power agents in real work environments.<br>Agent-centered training data<br>We developed Solar Open 2 with agentic use as a primary goal across both pre-training and post-training. Training scenarios cover search, tool calling (MCP), coding, and office work. The model learned not just to generate explanations, but to act in real environments, check the results, and correct itself when needed.<br>To build realistic agent training data, we developed our own data synthesis and verification pipeline. Starting from source data drawn from real work environments, we constructed settings where correct answers can be determined unambiguously, then repeated the process of selecting only the data that passed verification. Only this high-quality agent data went into training.<br>Search<br>Search agent data distinguishes between questions the model can answer with what it already knows and questions that require actual retrieval. Generated questions pass through multiple verification stages checking structure, non-triviality, searchability, solvability, and grounding in sources. This teaches the model to reach for search when it encounters knowledge it doesn't have.<br>Tool calling<br>Unlike conventional approaches that only judge whether the model selected the right tool, we have the model execute changes in a real environment, then read the changed state back to verify the result. The point is to separate generating a plausible answer from actually completing the task. This trains the model to take actions...