Vendor · Language models
Upstage has 7 models in the directory: 7 language models.
In the directory: 7 language models. 2 of them with open weights that can run locally.
Every model there ever was. Models with a profile lead to every benchmark, price and source.
| Solar Mini 4 | Sep 22, 2026 | Active | n/a | $0.10 / $0.40 | 115 | 512K | below 1 |
| Solar Pro 4 | Aug 11, 2026 | Active | n/a | $0.30 / $1.20 | 83 | 512K | 55 |
| Solar Pro 3 | January 2026 | Active | n/a | $0.15 / $0.60 | 145 | 131K | 29 |
| Solar Pro 2 | July 2025 | Active | n/a | $0.15 / $0.60 | n/a | 66K | 28 |
| Solar Mini | January 2024 | Active | n/a | $0.15 / $0.15 | n/a | 33K | 33 |
| SOLAR 10.7B Instruct v1.0 | Dec 14, 2023 | Active | n/a | n/a | n/a | 4K | 51 |
| SOLAR 10.7B v1.0 | December 2023 | Active | n/a | n/a | n/a | 4K | 29 |
Language model · Sep 22, 2026
Solar Mini 4 combines a 512K context window with an MoE model that activates just 3 of its 35 billion parameters per token. For local deployment, the provider specifies a single H100 with 80 GB of memory when using quantized weights.
Language model · Aug 11, 2026
Solar Pro 4 is designed for verifiable work products rather than short chat answers. Upstage names documents, terminal tasks, web research, and multi-step tool use as core cases. It offers a 512K-token context, up to 128K output tokens, and English, Korean, and Japanese support; Solar Pro 3 is the previous generation.
Language model · January 2026
Solar Pro 3 combines 102 billion total parameters with about 12 billion active per token and a 128K context. Upstage adds tool calling and stronger instruction following for multi-step work and reports roughly doubled agentic scores over Solar Pro 2, including 72.3 versus 36.0 on Tau2-all.
Language model · July 2025
Solar Pro 2 remains a dense 31-billion-parameter checkpoint that combines reasoning, tool use, and multilingual work with a 64K-token context. Upstage highlights multi-step and Korean use cases.
Language model · January 2024
Solar Mini stands out through depth up-scaling rather than a simple increase in parameter count. Upstage enlarged a 32-layer Llama 2 structure, then added continued pretraining plus instruction and alignment tuning. The result is a compact text checkpoint with 10.7 billion parameters.
Language model · Dec 14, 2023
SOLAR 10.7B Instruct v1.0 is the version of Upstage's SOLAR-10.7B-v1.0 fine-tuned for single-turn conversation, with 10.7 billion parameters. The base model is created through depth up-scaling, where Mistral 7B weights are integrated into the enlarged layers and pre-training then continues.
Language model · December 2023
SOLAR 10.7B v1.0 is Upstage’s compact open checkpoint, and its training method is the main reason to study it. Depth Upscaling changes the architecture of a 7B starting model, expands its layers, and continues pretraining with additional data. Upstage used the approach to target more capacity below the then-common 30B class without releasing an equally large dense checkpoint.
| Vendor | Models | Model types | Best language model | Top 10 | Open weights | New in 2026 |
|---|---|---|---|---|---|---|
| Anthropic | 33 | 33 language models | Claude Opus 5.5 (Rank 1 to 3) | 4 | 0 | 13 |
| OpenAI | 63 | 46 language models, 4 image models, 2 video models, 9 audio models, 2 specialist models | GPT-6 Astra (Rank 4 to 8) | 2 | 4 | 27 |
| 66 | 43 language models, 4 image models, 6 video models, 11 audio models, 2 specialist models | Gemini 4 Carbon (Rank 1 to 5) | 2 | 19 | 26 | |
| Meta | 32 | 29 language models, 1 image model, 1 video model, 1 specialist model | Muse Wonder Full (Rank 6 to 9) | 2 | 26 | 7 |
| DeepSeek | 23 | 23 language models | DeepSeek V4.1 Flash 1017 (Rank 13 to 20) | 0 | 23 | 9 |
| Moonshot AI | 12 | 12 language models | Kimi K3.1 Pro (Rank 10 to 13) | 0 | 10 | 7 |
| xAI | 16 | 12 language models, 2 image models, 2 video models | Grok 4.7 (Rank 18 to 24) | 0 | 1 | 12 |
| Upstage | 7 | 7 language models | not ranked yet | 0 | 2 | 3 |
By release date. Grey: retired models.
3 releases since January 2026, by release date.
Prices per 1M tokens for the current language models, according to Upstage or the provider linked in each model profile.
| Current model | Input | Output | Context |
|---|---|---|---|
| Solar Mini 4 | $0.10 | $0.40 | 512K |
| Solar Pro 3 | $0.15 | $0.60 | 131K |
| Solar Pro 2 | $0.15 | $0.60 | 66K |
| Solar Mini | $0.15 | $0.15 | 33K |
| Solar Pro 4 | $0.30 | $1.20 | 512K |
The directory lists 7 Upstage models: 7 language models. It counts every model Upstage ever released, including retired ones. 7 are currently available, the newest being Solar Mini 4, Solar Pro 4 and Solar Pro 3.
Solar Mini 4, released on Sep 22, 2026. The timeline below shows every release by date.
Through the API per 1M tokens (input / output): Solar Mini 4 $0.10 / $0.40. The prices of all 5 current language models are in the price table.
Yes, 2 of the 7 models have open weights. Each model profile lists how much memory it needs.
Solar Pro 3 with 145 tokens per second according to Artificial Analysis. A typical model manages 98.
Up to 512,000 tokens (≈ 770 A4 pages), which 2 of the current models reach, including Solar Mini 4. Every model's value is in the model list.
Data as of Nov 1, 2026