Solar Open — local AI model by Upstage
Written by Jakub Rusinowski · Last updated
Upstage's open-weight Solar models. Solar Open 2 is a 250B mixture-of-experts with 15B active per token whose hybrid attention keeps a KV cache on only a quarter of its layers, for a 1M-token context. Released under the Upstage Solar License: Apache-2.0 terms plus a duty to name derivative models 'Solar'.
Variants
The smallest Solar Open variant needs about 152 GB of VRAM at Q4_K_M — quantized weights plus framework overhead, before any KV cache.
| Model | VRAM at Q4 | VRAM | Context | Run it |
|---|---|---|---|---|
| Solar Open 2 250B-A15B → 250B (15B active) | ~151.9 GB | 1,048,576 | ollama run solar-open |
Memory is quantized weights plus overhead at Q4_K_M, from the same engine as the GPU & VRAM checker.
How to run Solar Open locally
Install Ollama, then pull the tag.
ollama run solar-openPick a size above for its own VRAM figure, speed estimate and install command.
Licence
Weights are downloadable and commercial use is permitted, subject to the licence’s acceptable-use terms.
Applies to: Solar Open 2 250B-A15BRecommended GPU
The cheapest catalogued GPU that runs Solar Open locally (min 152 GB VRAM) is the Apple M3 Ultra (512 GB).
Solar Open — frequently asked questions
How much VRAM does Solar Open need?
Solar Open needs about 152 GB VRAM at Q4_K_M quantization for its smallest variant. Variants: Solar Open 2 250B-A15B (152 GB, Q4_K_M). On Apple Silicon, unified memory counts toward this requirement.
Can I run Solar Open on an RTX 4090 (24 GB)?
Solar Open's smallest variant needs about 152 GB, which exceeds a single RTX 4090 (24 GB). Use multiple GPUs, a higher-VRAM card, or Apple Silicon with large unified memory.
What quantization should I use for Solar Open?
Q4_K_M is the best balance of quality and VRAM for Solar Open in most cases. Choose Q8_0 for near-lossless quality if you have spare VRAM, or smaller quants (Q3/Q2) only when memory is tight.
How do I run Solar Open with Ollama?
Solar Open has no local Ollama tag — the published tag is cloud-hosted, so running it sends your prompts to a hosted GPU rather than your own machine.