Qwen3.5 9B Base
Ministral 14B
- Context
- 4Ktokens
- Released
- 2y ago
- Size
- 10.7Bparameters
Upstage's SOLAR 10.7B Instruct v1.0 is an instruction-tuned version of the company's SOLAR-10.7B base model, which Upstage built with its depth up-scaling method by integrating Mistral 7B weights and continuing pre-training. Upstage's launch announcement billed Solar 10.7B as the world's top pre-trained LLM, citing a first-place ranking on the Hugging Face Open LLM Leaderboard. The instruct version was trained with supervised fine-tuning and direct preference optimization for single-turn conversation, and Upstage notes it is less suited to multi-turn chat.
How it compares
Size class
Small10.7BAn open-weight model of 5 to 25 billion parameters, the size that runs on a single consumer GPU.
Open models of a similar size from other labs
Vs. your stack
Add the models you use to see how this one compares.
Sign inAdditional source evidence
- Source model
- upstage/solar-10.7b-instruct-v1.0
- Source family
- solar
- Last updated
- 2023-12-12
- Model context
- 4K
- Model output limit
- 4K
- Open weights
- Yes
- File attachments
- No
- Reasoning
- No
- Tool calling
- No
- Temperature control
- Yes
Downloadable weights
Complete source record (JSON)
{
"id": "upstage/solar-10.7b-instruct-v1.0",
"name": "SOLAR 10.7B Instruct v1.0",
"description": "Upstage depth-upscaled 10.7B instruction model for single-turn chat",
"family": "solar",
"last_updated": "2023-12-12",
"attachment": false,
"reasoning": false,
"tool_call": false,
"temperature": true,
"modalities": {
"input": [
"text"
],
"output": [
"text"
]
},
"limit": {
"context": 4096,
"output": 4096
},
"release_date": "2023-12-12",
"open_weights": true,
"license": "CC-BY-NC-4.0",
"weights": [
{
"label": "Hugging Face",
"url": "https://huggingface.co/upstage/SOLAR-10.7B-Instruct-v1.0"
}
]
}Frequently asked questions
Running locally
8 GB at Q4, 13.6 GB at Q8, or 25.1 GB at FP16, estimated from its 10.7B parameters. All VRAM figures are estimates: weights plus about 15% and 0.5 GB for a short context.
Yes. A 24 GB or 16 GB GPU fits it at Q8 (estimated 13.6 GB), and a 12 GB or 8 GB GPU fits it at Q4 (estimated 8 GB).
Yes, at Q4 the estimated VRAM is 8 GB. Assuming macOS lets the GPU use about 75% of unified memory, the smallest common Mac memory size that fits is 16 GB.
Privacy and data
Yes, it has open weights, so you can self-host it for full control of your data.
Capabilities and specs
The context window is 4K tokens (4,096), and the maximum output is also 4K tokens (4,096).
No, it only accepts text. Images, audio, video and PDFs are not supported as inputs.
No, tool calling (function calling) is not supported.
No, a reasoning or thinking mode is not supported.
Quality and benchmarks
It ranks 379 on Arena Text (Arena's own rank) with a score of 1153.
The closest alternatives, matched on model kind and release date with no comparable price, are Tulu 2 DPO 70B by Ai2, OLMo 1B by Ai2 and Ferret by Apple.
Tulu 2 DPO 70BSOLAR 10.7B Instruct v1.0 vs Tulu 2 DPO 70BOLMo 1BSOLAR 10.7B Instruct v1.0 vs OLMo 1BFerretSOLAR 10.7B Instruct v1.0 vs Ferret
Licensing and access
It has open weights, released under the CC-BY-NC-4.0 license.
The weights can be downloaded from the Hugging Face repository upstage/SOLAR-10.7B-Instruct-v1.0.
Basics
SOLAR 10.7B Instruct v1.0 was made by Upstage and released on December 12, 2023.
Yes, it has open weights.
Generated from Vertical Tabs data, updated November 11, 2026.
Discussion
No comments yet
Be the first to share your thoughts.