Solar (Upstage)
Upstage's Korean-strong LLMs: Solar Pro 4, Mini 4 and open Solar Open 2
Overview
Solar is Upstage's LLM family, strong in Korean, English and Japanese: closed Solar Pro 4 (512K context) and Solar Mini 4 via an OpenAI-compatible API, and open-weight Solar Open 2 and Open 100B under the Apache-based Solar License, plus Document Parse. API from $0.10 per million tokens; prepaid plans from $100 a month.
Pricing
Pricing shown for reference only. These figures reflect RECATOOLS research as of 2 Oct 2026 and may be out of date or incomplete. This is not financial or purchasing advice — always confirm the current price on the provider’s official website before making any decision.
- +10% bonus credits monthly, +30% yearly
- Tier 1 rate limits
- Community support
- +15% / +35% bonus credits
- Tier 2 rate limits
- Email support, 3 business days
- +20% / +40% bonus credits
- Tier 3 rate limits
- Priority support, 1 business day
- Flexible rate limits
- Dedicated support
Use cases
What you can produce with Solar (Upstage)
- Solar Pro 4 flagship model (512K context)
- Solar Mini 4 low-cost model
- Open-weight Solar Open 2 and Open 100B
- Document Parse for structured document data
- OpenAI-compatible API and OpenRouter access
- Dedicated and on-premises deployment
Solar has matured into a credible model family, with a strong Korean-language flagship, a cheap small model and genuinely capable open weights that can be self-hosted commercially. Its benchmark wins are reported by Upstage, older models retire at the end of October so integrations need updating, and language coverage beyond Korean, English and Japanese is limited. A strong choice for Korean and Japanese workloads; compare against global models for other languages.
What we found
Solar is the family of large language models from Upstage, a Seoul AI company, built with particular strength in Korean, English and Japanese. The current line-up is Solar Pro 4, a closed flagship released in August 2026 with a 512,000-token context and up to 128,000 tokens of output; Solar Mini 4, a smaller mixture-of-experts model with 3 billion active parameters released in October; and open-weight models — Solar Open 2, a 250-billion-parameter mixture-of-experts model with 15 billion active and a million-token context, released in July 2026, and the earlier Solar Open 100B. The open weights use Upstage's Solar License, which follows Apache 2.0 and permits commercial use but requires derivatives to carry the Solar name. Upstage reports Solar Pro 4 scoring 42 on the Artificial Analysis index ahead of several rivals and heavy early use on OpenRouter; those are its own figures. Solar Pro 3, Pro 2 and the old Mini are retired on 30 October 2026. Upstage also sells Document Parse for turning documents into structured data, and models are available through an OpenAI-compatible API, SolarChat, OpenRouter and dedicated or on-premises deployment.
Upstage Co., Ltd. was founded in Seoul in 2020 by Sung Kim and became South Korea's first generative-AI start-up unicorn in April 2026 with a Series C valuing it above ₩1 trillion; it then acquired AXZ, operator of the Daum portal, from Kakao and is preparing for a listing. Its revenue rose from ₩5.9 billion in 2022 to ₩24.8 billion in 2025. In Southeast Asia, it co-developed a Thai sovereign language model for KT's project with Jasmine Technology Solution.
API pricing is per million tokens: Solar Mini 4 costs $0.10 in and $0.40 out, and Solar Pro 4 $0.30 in and $1.20 out, both on launch discounts. Document Parse costs $0.01 to $0.03 a page. Prepaid plans add bonus credits and higher rate limits: Explore from $100 a month, Build from $500 and Scale from $5,000, with Enterprise custom. The open weights are free to download.
Solar suits teams needing strong Korean and Japanese language models, open or hosted.
Compiled by RECATOOLS from public sources: vendor documentation and pricing, code repositories and published reporting.
Notable facts
- Solar 10.7B was trained by a South Korean company and outperformed all European and American models of similar size, demonstrating APAC's growing contributions to frontier AI.
- The Depth Upscaling technique used for Solar allows models to be built more efficiently by combining and extending existing pre-trained architectures.
- Solar was the first non-Western model to achieve #1 on the HuggingFace Open LLM Leaderboard.
Frequently asked questions
About this listing
This entry was compiled from publicly available data including Solar (Upstage)'s official website, press releases, documentation, and reputable third-party publications. RECATOOLS is not affiliated with Solar (Upstage) unless explicitly stated.
Third-party AI tools update their pricing, features, availability, and policies frequently. Information here may be outdated by the time you read this — we make reasonable efforts to keep listings current, but cannot guarantee absolute accuracy.
For the latest details, please refer to Solar (Upstage) directly →
Spotted something out of date? Suggest an update →
Related reading
Cybersecurity
CISA Orders Three-Day Fix for Exploited Collabora Flaw in Nextcloud's Built-In Office Serv...
AI & ML
DeepSeek Open-Sources DeepInfer, the Engine That Serves Its V4.1 Flash Model on NVIDIA and...
AI & ML
A Chinese AI Hacking Tool and Claude Code Were Behind Korea's Bank Breaches, CrowdStrike S...
Alternatives to Solar (Upstage)
More in LLMs & Chat
