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. | 1.5B/7B/72B | 0.5B/1.5B/3B/7B/14B/32B/72B |
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. | 4k | 128k |
| API pricingUSD per 1M tokens · lower wins | ||
Cheapest inputLowest input rate across every provider on OpenRouter, the lab included. The cheapest endpoint may run a quantised build or a shorter context — see "Available from" on the model page. | — | $0.288DeepInfra |
Cheapest outputLowest output rate across every provider on OpenRouter, the lab included. May come from a different provider than the cheapest input. | — | $0.32DeepInfra |
| — | $92.00DeepInfra · fp8 | |
| Benchmarks | ||
MATHReasoning — accuracy on the original competition mathematics test set | 84% | 83.1% |
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. | 96.7% | 95.8% |
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. | — | 55.5% |
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 | — | 84.1% |
MMLU-ProKnowledge — accuracy on the ten-option MMLU-Pro benchmark | — | 71.1% |
| Overview | ||
| Company | Qwen | Qwen |
| Release date | Aug 8 2024 | Sep 19 2024 |
| Access | Open Weight | Open Weight |
| Model details | View model | View model |
Other comparisons
Qwen2-MathvsClaude Fable 5.2Qwen2.5vsClaude Fable 5.2Qwen2-MathvsGPT-6.2 AstraQwen2.5vsGPT-6.2 AstraQwen2-MathvsGemini 4 NeonQwen2.5vsGemini 4 NeonQwen2-MathvsMuse EmberQwen2.5vsMuse EmberQwen2-MathvsGrok 4.7Qwen2.5vsGrok 4.7Qwen2-MathvsDeepSeek-V4.1-Flash-1017Qwen2.5vsDeepSeek-V4.1-Flash-1017Frequently asked questions
Qwen2-Math leads Qwen2.5 on 2 of the 2 benchmarks they both report (MATH, GSM8K). Qwen2-Math shipped 42 days before Qwen2.5, so benchmark comparisons should account for the intervening progress.
Qwen2-Math has 1.5B/7B/72B parameters, while Qwen2.5 has 0.5B/1.5B/3B/7B/14B/32B/72B. Context windows are 4k (Qwen2-Math) vs 128k (Qwen2.5).
On MATH, Qwen2-Math leads at 84% vs Qwen2.5 at 83.1%. On GSM8K, Qwen2-Math leads at 96.7% vs Qwen2.5 at 95.8%.