Solar Open 2
Solar Open 2
A large open model built specifically to be an agent rather than a chatbot.
Updated
The read
Upstage tuned Solar Open 2 for multi-step autonomous work: planning, tool use, recovering from its own mistakes, and running long chains without losing the thread. It is a large sparse model with a small active parameter count, published openly, and it is one of the few open releases where agent behaviour was the design goal rather than an emergent bonus. Korean and English are its strong languages.
Where it shines
- Tuned for autonomous multi-step work, not just chat
- Sparse design keeps inference cost reasonable for its size
- Among the strongest open models for Korean
Worth knowing first
- Self-hosting is the only realistic route: no large hosted ecosystem
- Weaker outside Korean and English
Index score
79.2
Provisional. This score is read from the product's public capability, not from the one-prompt rebuild the fully reviewed entries went through, so treat it as a placing rather than a verdict.
- Rank
- 27 of 53
- Builds
- LLMs
- Category
- Open weights
- Output
- Published open weights, agent-tuned
- Pricing
- Free tier
- Best for
- Self-hosted agents that run long autonomous chains
Scores are our own editorial judgement, weighted craft 55, speed 10, control 15 and value 20.
Also worth a look
Full index- 05Kimi K3Kimi K3ProvisionalThe largest open-weights model anyone has shipped, and it reads a million tokens without blinking.Open weightsFree tier87.5
- 06GLM-5.2GLM-5.2ProvisionalFrontier-class output, an MIT licence, and a million tokens of context. That combination is the story.Open weightsFree tier87.1
- 12Qwen 3.6Qwen 3.6ProvisionalThe open Qwen release most people should actually run, in sizes that fit real hardware.Open weightsFree tier83.3
