Compare AI models
| Specifications | ||
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. | 2M | 1M |
| Benchmarks | ||
LiveCodeBenchCompetitive coding — Coding problems published so recently the AI can't have seen them in training — a contamination-free test of raw programming skill. Higher is better. | 34.3% | 37.8% |
MATHReasoning — accuracy on the original competition mathematics test set | 86.5% | 81.1% |
MMLU-ProKnowledge — accuracy on the ten-option MMLU-Pro benchmark | 75.8% | 64.5% |
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. | — | 86.6% |
IFEvalAgentic tasks — following verifiable instructions in generated responses | — | 76.3% |
GPQA DiamondScience — Graduate-level science questions in biology, physics, and chemistry — hard enough that subject-matter PhDs score around 65%. Higher is better. | 59.1% | — |
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. | — | 93.8% |
MMMUMultimodal — Tests the AI on understanding images and text together across many college subjects. Higher is better. | 65.9% | — |
| Overview | ||
| Company | Qwen | |
| Release date | Sep 24 2024 | Nov 15 2024 |
| Access | Closed | Closed |
| Model details | View model | View model |
Other comparisons
Gemini 1.5 Pro-002vsClaude Fable 5.2Qwen2.5-Turbo (1M)vsClaude Fable 5.2Gemini 1.5 Pro-002vsGPT-6.2 AstraQwen2.5-Turbo (1M)vsGPT-6.2 AstraGemini 1.5 Pro-002vsMuse EmberQwen2.5-Turbo (1M)vsMuse EmberGemini 1.5 Pro-002vsGrok 4.7Qwen2.5-Turbo (1M)vsGrok 4.7Gemini 1.5 Pro-002vsDeepSeek-V4.1-Flash-1017Qwen2.5-Turbo (1M)vsDeepSeek-V4.1-Flash-1017Gemini 1.5 Pro-002vsMistral Large 4Qwen2.5-Turbo (1M)vsMistral Large 4Frequently asked questions
Gemini 1.5 Pro-002 leads Qwen2.5-Turbo (1M) on 2 of the 3 benchmarks they both report (LiveCodeBench, MATH, MMLU-Pro). Gemini 1.5 Pro-002 shipped 52 days before Qwen2.5-Turbo (1M), so benchmark comparisons should account for the intervening progress.
Context windows are 2M (Gemini 1.5 Pro-002) vs 1M (Qwen2.5-Turbo (1M)).
On LiveCodeBench, Qwen2.5-Turbo (1M) leads at 37.8% vs Gemini 1.5 Pro-002 at 34.3%. On MATH, Gemini 1.5 Pro-002 leads at 86.5% vs Qwen2.5-Turbo (1M) at 81.1%. On MMLU-Pro, Gemini 1.5 Pro-002 leads at 75.8% vs Qwen2.5-Turbo (1M) at 64.5%.