Bili Hardcore Download

AI-powered auto quiz solver for Bilibili hardcore membership exam. Supports DeepSeek, Gemini and OpenAI models. Calls Bilibili API directly with QR login for automated question answering.

⭐ 1,456 stars on GitHub
Latest Release: v0.5.2

About Software

Bili Hardcore is an automated quiz solver for Bilibili's hardcore membership exam. It uses LLM APIs to intelligently answer questions by calling Bilibili's API directly. Supports DeepSeek V3.1, Gemini 2.5 Flash, and OpenAI-compatible endpoints. Requires level 6 Bilibili account with 3 daily attempts allowed.

Login via QR code scan, choose quiz category, complete captcha verification and let the tool handle 100 questions automatically. Config and credentials saved to ~/.bili-hardcore for reuse. Available as executables for Windows, macOS and Ubuntu, or run from Python 3.9+ source. Licensed under MIT.

Use Cases:

  • Automate Bilibili hardcore membership exam with AI-powered answers
  • Use DeepSeek, Gemini or OpenAI-compatible models for quiz solving
  • Scan QR code to login and complete up to 100 questions per session
  • Save API keys and login info locally for repeated use
  • Run on Windows, macOS or Ubuntu with standalone executables

Downloads

v0.5.2 September 05, 2025
bili-hardcore-windows-latest-v0.5.2.exeexe
v0.5.1 August 04, 2025
bili-hardcore-windows-latest-v0.5.1.exeexe
v0.5.0 April 24, 2025
bili-hardcore-windows-latest-v0.5.0.exeexe
v0.4.3 April 17, 2025
bili-hardcore-windows-latest-v0.4.3.exeexe
v0.4.2 April 12, 2025
bili-hardcore-windows-latest-v0.4.2.exeexe
v0.4.1 April 10, 2025
bili-hardcore.exeexe
v0.4.0 April 07, 2025
bili-hardcore_x86_64.exeexe
v0.3.1 April 07, 2025
bili-hardcore_x86_64.exeexe
v0.3.0 April 07, 2025
bili-hardcore.exeexe
v0.2.0 April 06, 2025
bili-hardcore.exeexe
v0.1.2 April 06, 2025
bili-hardcore.exeexe
v0.1.1 April 06, 2025
bili-hardcore.exeexe
v0.1.0 April 06, 2025
bili-hardcore.exeexe
v0.0.2 April 05, 2025
bili-hardcore.exeexe
v0.0.1 April 04, 2025
bili-hardcore.exeexe

Package Info

Last Updated
Sep 05, 2025
Latest Version
v0.5.2
License
MIT
Total Versions
15

README

Bili-Hardcore

B 站硬核会员自动答题工具,利用 LLM 实现智能答题功能。

可用的模型:

  • DeepSeek (V3.1)
  • Gemini (gemini-2.5-flash)(为了防止风控,答题间隔5秒,速度较慢,建议使用其他模型)
  • OpenAI 风格的其他 API(OpenAI、火山引擎、硅基流动等,可自定义 url、模型名称)

不建议思考模型,思维链过长可能会导致超时请求失败

使用前须知

  • 程序仅在本地调用 B 站 API 和 LLM API,不会上传您的登录信息和 API Key,请放心使用
  • 请确保您的 B 站账号已满 6 级,根据 B 站规则,6 级用户才可以进行硬核会员试炼
  • 硬核会员试炼每天有 3 次答题机会(指的是100道题全部提交,显示答题结果后,或在 B 站 APP 手动结束答题),达到限制后需要 24 小时后才能重新答题,具体时间可以前往 B 站 APP 答题页面查看
  • 首次输入模型配置和登录后,会将信息保存到 ~/.bili-hardcore,下次运行时会自动读取。如配置错误或遇到奇怪问题,运行脚本后请选择“重新开始”
  • 没有 API Key 的可以自己去免费去硅基流动注册一个账号,会送 14 元免费额度,这是我的邀请链接 (https://cloud.siliconflow.cn/i/9Fur0aVC)
  • 如果不知道 API 基础 URL 和模型名称如何填写,可以参考配置填写示例
  • 请合理使用,遵守 B 站相关规则

使用说明

方式一:从 release 下载可执行文件

Windows

  1. 下载 bili-hardcore-windows-*.exe
  2. 双击 exe 运行或在命令行中执行 .\bili-hardcore-windows-*.exe

Mac

  1. 下载 bili-hardcore-macos-*
  2. 命令行中执行 chmod +x bili-hardcore-macos-* && ./bili-hardcore-macos-*

    如果碰到"Apple could not verify xxx is free of malware that may harm your Mac"问题,可以在系统设置 > 隐私与安全性中点击仍要打开,或者命令行执行 xattr -d com.apple.quarantine ./bili-hardcore-macos-*

Ubuntu

  1. 下载 bili-hardcore-ubuntu-*
  2. 命令行中执行 chmod +x bili-hardcore-ubuntu-* &&./bili-hardcore-ubuntu-*

方式二:从源码运行

请使用 Python 3.9 及以上版本运行

  1. 克隆项目到本地
git clone [项目地址]
cd bili-hardcore
  1. 安装依赖
pip install -r requirements.txt
  1. 运行主程序
python bili-hardcore/main.py

使用流程

  1. 选择回答模型
  2. 输入自己的 API Key
  3. 扫描二维码登录
  4. 输入要进行答题的分类
  5. 查看并输入图形验证码
  6. 程序会自动开始答题流程

常见问题

  1. 二维码乱码:请尝试使用其他命令行工具运行,或手动生成二维码进行扫码
  2. 答题不及格:尝试使用历史分区答题,历史分区的准确率较高
  3. AI 卡在一个问题一直过不去,回复类似于“无法确认、我不清楚”:换一个模型,或者去 B 站 APP 手动把卡住的题目过了,切记不要在 B 站答题页面点击左上角返回按钮退出,会结束答题
  4. DeepSeek 官方模型在答题过程中报 400 错误:检查一下题目里是否有敏感词,如果有敏感词,需要去 B 站 APP 手动过这一题,或者换个模型回答
  5. 获取分类失败,41099 错误:请前往 B 站 APP 答题页面确认是否已达到答题限制

Gemini 模型使用问题及解决办法

  1. 答题触发 429 错误:应该是触发了 Gemini 每分钟调用限制或触发了风控,依次尝试以下操作:
    1. 可以稍等一下重新运行,会接着中断的题目继续回答
    2. 如果还不行,尝试切换节点(修改IP)
    3. 再不行就需要手动修改一下代码里的 prompt
    4. 终极解决办法:别用 Gemini 模型了,用 DeepSeek 模型
  2. 开始答题直接之后软件直接退出:需要切换到大陆及香港以外的节点进行答题

运行截图

!PixPin_2025-04-08_15-45-29 (https://github.com/user-attachments/assets/70b3930c-c60f-43f7-8d82-c5225997ebc5)

Related Software

Obamify

Desktop and web application using advanced pixel reassignment algorithm to transform any image into Obama or other target images with adjustable resolution, animation control, and mathematical optimization options.

⭐ 1,733

Cagent

Docker's multi-agent runtime for building and orchestrating specialized AI agents defined in YAML, with MCP tool integration, smart delegation, RAG retrieval, and support for multiple LLM providers including OpenAI, Anthropic, and Gemini.

⭐ 1,732agents, ai

Athas

Lightweight cross-platform code editor built with Tauri combining Rust and React, featuring external editor support for Neovim and Helix, Git integration, AI agents, Vim keybindings, and integrated terminal for Windows, macOS, and Linux.

⭐ 1,722editor, ide, athas

Audio Visual

Cross-platform video parsing tool built with Electron for streaming content from Chinese platforms like Tencent Video, iQiyi, Youku, Bilibili, and Mango TV with multiple parser interfaces, modern UI, automatic updates, and support for Windows, macOS, and Linux.

⭐ 1,716

Nook

Fast, minimal macOS browser with sidebar-first vertical tab design, built with SwiftUI using modern macOS APIs for high performance and low memory usage, featuring extension support, profile isolation, split-view tabs, and command palette.

⭐ 1,677

Klee

AI-powered knowledge management desktop app with RAG capabilities, Tiptap collaborative editor, and dual modes: Cloud sync via Supabase or complete offline privacy with local Ollama AI, SQLite storage, and LanceDB vector search.

⭐ 1,676app, klee, llamaindex