Model

Qwen1.5-110B

OPEN

Alibaba · Qwen1.5 family · released Apr 25, 2024

First 100B+ dense Qwen1.5, rivaled Llama-3-70B base; announced alongside the closed Qwen-Max-0428.

ReasoningCodingVisionFunction callingTool useAgentic

665.5

Elo · rank #271

Parameters

111B

Active params

111B (dense)

Context

32K tokens

Architecture

Dense decoder-only Transformer (GQA)

License

Tongyi Qianwen License Agreement

Languages

—

API price (in/out)

No hosted API

Modalities

text

Benchmark results

Bar shows position within the tracked field; marker = field best

AlpacaEvalHuman preference43.9%#23

best: Tülu 2+DPO 70B · 95.1%

ARC-ChallengeReasoning69.6%#49

best: Llama 3.1 405B · 96.9%

BIG-Bench HardReasoning74.8%#44

best: ERNIE 4.5 300B-A47B · 94.3%

GPQA DiamondReasoning35.9%#239

best: GPT-5.6 · 94.6%

GSM8KMath85.4%#65

best: Llama 3.1 405B · 96.8%

HellaSwagReasoning87.5%#18

best: Claude 3 Opus · 95.4%

HumanEvalCoding52.4%#117

best: Claude Opus 4.5 · 99.4%

IFEvalReasoning34.2%#138

best: Gemma 4 26B A4B · 98.5%

MATH-500Math49.6%#136

best: GPT-5 · 99.4%

MBPPCoding58.1%#60

best: Llama-3.3-Nemotron-Super-49B v1 (Reasoning On) · 91.3%

MMLU-ProKnowledge48.5%#134

best: Claude Fable 5 · 91.5%

MMLUKnowledge80.4%#88

best: OpenAI o3 · 92.9%

MT-BenchHuman preference8.88#15

best: Hunyuan-Large (A52B) · 9.4

Run it locally

VRAM @ Q4

65 GB

VRAM @ FP16

222 GB

Fits on (Q4)

M3 Max 128GBM3 Ultra 512GBA100 80GBH100 80GBH200 141GBB200 192GB

Exceeds a single RTX 4090's 24GB VRAM even at Q4; requires multi-GPU or CPU offload.

Quantizations

GGUF · AWQ · GPTQ

Qwen1.5 family

Elo progression across releases

Qwen1.5-72BFeb 2024545.1 Qwen1.5-MoE-A2.7BMar 202494.1 CodeQwen1.5-7BApr 2024526.5 Qwen1.5-110BApr 2024665.5

Resources

HFQwen/Qwen1.5-110B-Chat GHQwenLM/Qwen1.5 DOCSqwenlm.github.io/blog/qwen1.5-110b

API price weights · each benchmark row carries its own source badge (see methodology)