Model
Model explorer

Upstage Solar Pro

OPEN
Upstage · SOLAR family · released Sep 11, 2024

Sept 2024 MIT-licensed preview (4K ctx, English) topped the <70B Open LLM Leaderboard; a Nov 2024 official update added 32K context and multilingual support via API only.

ReasoningCodingVisionFunction callingTool useAgentic
747.6
Elo · rank #259
Parameters
22B
Active params
22B (dense)
Context
4K tokens
Architecture
Depth Up-Scaled (DUS) dense transformer built from Phi-3-medium (14B)
License
MIT
Languages
—
API price (in/out)
No hosted API
Modalities
text
Benchmark results
Bar shows position within the tracked field; marker = field best
ARC-ChallengeReasoning68.9%#53
best: Llama 3.1 405B · 96.9%
BIG-Bench HardReasoning67.3%#66
best: ERNIE 4.5 300B-A47B · 94.3%
GPQA DiamondReasoning36.4%#236
best: GPT-5.6 · 94.6%
GSM8KMath89.7%#43
best: Llama 3.1 405B · 96.8%
HellaSwagReasoning86.4%#29
best: Claude 3 Opus · 95.4%
IFEvalReasoning84.4%#68
best: Gemma 4 26B A4B · 98.5%
MBPPCoding61.6%#55
best: Llama-3.3-Nemotron-Super-49B v1 (Reasoning On) · 91.3%
MMLU-ProKnowledge52.1%#130
best: Claude Fable 5 · 91.5%
MMLUKnowledge79.1%#97
best: OpenAI o3 · 92.9%
Run it locally
VRAM @ Q4
13 GB
VRAM @ FP16
44 GB
Fits on (Q4)
RTX 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
SOLAR family
Elo progression across releases
Upstage SOLAR 10.7BDec 2023396.9Upstage Solar ProSep 2024747.6Upstage Solar Pro 2 (Chat)Jul 20251103.4Upstage Solar Pro 2 (Reasoning)Jul 20251419.4
Resources
HFupstage/solar-pro-preview-instructDOCSupstage.aiAPIAvailable via Upstage API
API price weights · each benchmark row carries its own source badge (see methodology)