Qwen2.5-MathOpen Weight

Released

Qwen2.5-Math is an AI model released by Qwen on Thursday, Sep 19 2024, the same day as Qwen2.5-Coder. It is an open-weight model — the trained weights are available to download and run. It was released in 1.5B and 7B and 72B parameter sizesThe 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., with a 4k token 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.. Benchmark results (shown below) cover MATH, and GSM8K.

Benchmarks

MATHReasoning — accuracy on the original competition mathematics test set
85.9%
#6 of 60
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.
95.9%
#4 of 56

Compare Qwen2.5-Math with

Qwen2.5-Math

About

Qwen2.5-Math arrived on September 19, 2024 with downloadable 1.5B, 7B and 72B checkpoints. It expanded the earlier specialist family with bilingual mathematical reasoning and optional use of a Python interpreter during problem-solving.

In Qwen's technical report, the 72B instruction checkpoint scored 85.9 on MATH with a single chain-of-thought answer. The team also published separate results for tool-integrated reasoning, majority voting and reward-model selection.

Frequently asked questions

Qwen2.5-Math was released by Qwen on Thursday, Sep 19 2024.

All Qwen releases

42 tracked