Model
Model explorer

Upstage SOLAR 10.7B

OPEN
Upstage · SOLAR family · released Dec 13, 2023

DUS-merged from two Mistral-7B copies; the Instruct variant (H6=74.20) briefly topped the Dec 2023 HF Open LLM Leaderboard.

ReasoningCodingVisionFunction callingTool useAgentic
396.9
Elo · rank #331
Parameters
10.7B
Active params
10.7B (dense)
Context
4K tokens
Architecture
Depth Up-Scaled (DUS) dense transformer — two 24-layer Mistral-7B-derived stacks merged to 48 layers, then continually pretrained
License
Apache 2.0 (base); Instruct variant is CC-BY-NC-4.0
Languages
API price (in/out)
No hosted API
Modalities
text
Benchmark results
Bar shows position within the tracked field; marker = field best
ARC-ChallengeReasoning62.0%#82
best: Llama 3.1 405B · 96.9%
GSM8KMath55.5%#124
best: Llama 3.1 405B · 96.8%
HellaSwagReasoning84.6%#45
best: Claude 3 Opus · 95.4%
MMLUKnowledge65.5%#175
best: OpenAI o3 · 92.9%
TruthfulQAKnowledge45.0%#66
best: Phi-3.5-MoE (16x3.8B, 6.6B active) · 77.5%
WinoGrandeReasoning83.7%#15
best: PaLM 2 · 90.9%
Run it locally
VRAM @ Q4
7 GB
VRAM @ FP16
21 GB
Fits on (Q4)
RTX 3060 12GBRTX 4070 Ti 16GBRTX 3090 24GBRTX 4090 24GBRTX 5090 32GBM4 Pro 48GBM3 Max 128GBM3 Ultra 512GBA100 80GBH100 80GBH200 141GBB200 192GB
Throughput data unavailable.
Quantizations
GGUF Q4 · AWQ · GPTQ
API price weights · each benchmark row carries its own source badge (see methodology)