Compare AI models
| Specifications | ||
ParametersThe learned values that shape how a model responds. B means billion and T means trillion. For mixture-of-experts models, the total count is larger than the number active for each token. Parameter count alone does not measure quality. | 1.8B/3.25B | 110B |
Context windowHow much information the model can work with in one request, measured in tokens — pieces of words or other data. This includes your question, attachments, conversation and reply. Limits can vary by model variant and provider. | 32k | 32k |
| Benchmarks | ||
MATHReasoning — accuracy on the original competition mathematics test set | 22.8% | 49.6% |
MMLUGeneral knowledge — A 57-subject multiple-choice exam — history, law, medicine, maths — that was the standard measure of how much a model knows from 2020 until roughly 2024, when frontier scores crowded into the high 80s and labs moved on to harder tests. The scores here were published years apart under different testing setups, so read them as a historical record rather than a like-for-like ranking. Higher is better. | 55.8% | 80.4% |
HumanEvalFunction synthesis — 164 small Python problems: the AI is given a function's description and has to write the working function. This was the coding benchmark of the GPT-3.5 and GPT-4 era, before the field moved to fixing real bugs in real repositories. Higher is better. | — | 52.4% |
HellaSwagReasoning — commonsense completion of short scenarios | — | 87.5% |
GSM8KGrade-school math — Grade-school maths word problems that take a few steps of arithmetic to work through. It separated the models of 2022 and 2023 sharply, then saturated. One caveat on the historical numbers: OpenAI included part of the GSM8K training set in GPT-4's pre-training mix, so GPT-4's score is not a clean few-shot result. Higher is better. | — | 85.4% |
MathVistaMultimodal reasoning — mathematical reasoning about visual information | 30.6% | — |
MMMUMultimodal — Tests the AI on understanding images and text together across many college subjects. Higher is better. | 32.6% | — |
| Overview | ||
| Company | Qwen | |
| Release date | Dec 6 2023 | Apr 25 2024 |
| Access | Closed | Open Weight |
| Model details | View model | View model |
Other comparisons
Gemini 1.0 NanovsClaude Fable 5.2Qwen1.5-110BvsClaude Fable 5.2Gemini 1.0 NanovsGPT-6.2 AstraQwen1.5-110BvsGPT-6.2 AstraGemini 1.0 NanovsMuse EmberQwen1.5-110BvsMuse EmberGemini 1.0 NanovsGrok 4.7Qwen1.5-110BvsGrok 4.7Gemini 1.0 NanovsDeepSeek-V4.1-Flash-1017Qwen1.5-110BvsDeepSeek-V4.1-Flash-1017Gemini 1.0 NanovsMistral Large 4Qwen1.5-110BvsMistral Large 4Frequently asked questions
Qwen1.5-110B leads Gemini 1.0 Nano on 2 of the 2 benchmarks they both report (MATH, MMLU). Gemini 1.0 Nano shipped 141 days before Qwen1.5-110B, so benchmark comparisons should account for the intervening progress.
Gemini 1.0 Nano has 1.8B/3.25B parameters, while Qwen1.5-110B has 110B. Context windows are 32k (Gemini 1.0 Nano) vs 32k (Qwen1.5-110B). Gemini 1.0 Nano is closed, while Qwen1.5-110B is open weight.
On MATH, Qwen1.5-110B leads at 49.6% vs Gemini 1.0 Nano at 22.8%. On MMLU, Qwen1.5-110B leads at 80.4% vs Gemini 1.0 Nano at 55.8%.