Skip to content
Parameters 7BApache-2.0Estimated

Qwen2.5-7B

拿来微调出自己下游模型的 7B 基座权重。

Official source:https://huggingface.co/Qwen/Qwen2.5-7B阿里云通义千问(Qwen)You are leaving AIWorksWell for 阿里云通义千问(Qwen). That site is not under our control; its content, privacy and security are its own responsibility. Please judge for yourself.

What hardware it needs

Estimated
Parameters
7B
Architecture
Dense
Context
128K
License
Apache-2.0
Min VRAM
6G
RAM (CPU-only)
8G
Download size
—G

Requirements by quantisation

The same model at different quantisations is a different download — the most common misjudgement in local AI.

Suggested quantisationSizeMin VRAMQuality
F1628G32GOriginal precision
Q8_014.8G17GPractically lossless
Q6_K11.5G13GNear lossless
Q5_K_M9.9G11GVery good
Q4_K_M★8.5G10GBest value for most people
Q3_K_M7G8GNoticeable loss
Q2_K5.3G6.5GLast resort only

Method:显存为 Q4_K_M 档参考

Your device

A plain-language profile of every model you can run yourself: parameter count, hardware it needs, which quantisation fits. No rankings, no prices — just what it takes to run.

Save your device profile once and this page will tell you whether it runs on your machine.Add your device·No account? Try a one-off check

Install prompt

Copy it, paste it into your AI (Claude Code, Codex, ChatGPT…), and it installs and verifies for you.

45 lines
我需要你帮我在我的机器上安装并运行一个软件。

【软件信息】
名称:Qwen2.5-7B
用途:拿来微调出自己下游模型的 7B 基座权重。
官方来源:https://huggingface.co/Qwen/Qwen2.5-7B
许可:Apache-2.0
Scroll inside the box with your mouse wheel — copying still gives you the full prompt1238 chars

Copy → paste into your own AI → it installs for you

About this entry

它是什么

Qwen2.5-7B 是这一代 7B 系列的基座权重:只做过预训练、没有指令微调,官方模型卡自己写着「不建议用基座模型对话」。它该被当成微调的起点,而不是一个能直接用的助手。来源:https://huggingface.co/Qwen/Qwen2.5-7Bhttps://huggingface.co/Qwen/Qwen2.5-7B

怎么用起来

拉 transformers 格式的权重做 SFT / LoRA;只想先跑通推理链路,就拿社区转好的 GGUF 量化版(Q4_K_M 约 6GB 显存)。要对话请换指令微调版。来源:https://huggingface.co/Qwen/Qwen2.5-7Bhttps://huggingface.co/Qwen/Qwen2.5-7B

以上坑来自 https://huggingface.co/Qwen/Qwen2.5-7Bhttps://huggingface.co/Qwen/Qwen2.5-7B 与官方文档。

来源与署名

Mentioned from:https://huggingface.co/Qwen/Qwen2.5-7Bhttps://huggingface.co/Qwen/Qwen2.5-7B · 原作者 阿里云通义千问(Qwen)(Apache-2.0)。

ProvenanceVerified

  1. OriginalCurator2026-10-19Evidence这一条就是原作(基座权重的发布方)。上游 README 第 23 行原话 "This repo contains the base 7B Qwen2.5 model";第 37 行 "We do not recommend using base language models for conversations"。★ 为什么值得单独收一条:站内的 Instruct 与 VL 的 `base_model` 都指向它,而此前站内没有它 ⇒ 那两条只能写「根在站外」,族树永远长不出「一个根、两个下游」的形状(D413 / 19 #75)。

The provenance chain records claims only: "Verified" requires evidence; everything else is a declaration. We do not vouch for any attribution (41 §4).