其中一个已经不复存在
Google 将 Gemini CLI 退役，并入 Antigravity CLI，面向消费者的访问已于 2026 年 6 月 18 日停止处理请求。我们保留这一行，因为知道它被什么取代，比悄悄删掉它更有用。
编程框架
编码智能体不是模型
您根据能力选择模型，根据工作完成方式选择框架。这些是 2026 年值得了解的框架，每项声明都有来源和日期——包括那个已经退役的。
对比它们所有行均于 2026年8月14日 验证
请先阅读
框架是将模型变成智能体的客户端：它负责循环、工具、文件访问和权限。两个人通过不同框架使用同一模型，会得到截然不同的结果——并且支付截然不同的费用，因为框架决定了这项工作消耗多少 token。
实验室现在通过框架进行基准测试
当 Z.ai 发布 GLM-5.3 的分数时，它是在 Claude Code——竞争对手的框架——中运行的。这说明框架层已成为标准基础设施，也说明模型公布的分数部分反映了它是在哪个客户端中测得的。
对比这些框架
按每个框架处理模型的方式筛选，然后选择一行查看其完整详情、限制和来源。已退役的框架默认隐藏。
| 框架 | 模型 | 许可证 | 运行于 | 定价 |
|---|---|---|---|---|
| can1357 (community) | 任意模型Any provider via BYOK — Anthropic, OpenAI, Google, xAI, DeepSeek, Mistral, plus local servers (Ollama, llama.cpp, vLLM) | 开源 | Terminal, Zed, Node.js SDK, One-shot CLI | Free, MIT-licensed. You pay only your own provider API bills; no subscription and no vendor account required. |
| Anthropic | 锁定到单一供应商Anthropic models only (Claude Fable, Opus, Sonnet, Haiku) | 专有 | Terminal, VS Code, JetBrains, Desktop, Web, GitHub, Slack | Included with Claude paid plans; also usage-billed via the Anthropic API |
| Anomaly | 任意模型200+ providers through Models.dev, including local models; log in with an existing ChatGPT Plus or Pro or GitHub Copilot subscription, or use OpenCode Go | 开源 | Desktop app (macOS, Windows, Linux), Terminal, Mobile (iPhone, via TestFlight), SDK | Free and MIT-licensed, with free models included and no account needed. OpenCode Go is $10/month and Go Plus is $40/month; extra credit can be topped up. |
| Z.ai | 自有模型，但不锁定GLM family (GLM-5.3, GLM-5.2), including a self-hosted GLM endpoint | 专有 | Desktop (macOS, Windows, Linux), Mobile steering via WeChat / Feishu | Via the GLM Coding Plan's points quota — off-peak calls cost 50% of standard points |
| Command Code | 任意模型Anthropic, OpenAI, Google, xAI, DeepSeek, Qwen, Kimi, GLM, MiniMax and more; built around open models, with DeepSeek V4.1 Flash as the default in its latest changelog entry | 专有 | Terminal (interactive and headless), Desktop app (macOS, Linux, Windows), Background sandbox | Go $1/month plus a processing fee (includes $10 of credits); GOAT $10; Pro $20; Max $100 or $200; Teams $40; Provider API plan $15/month then pay as you go; Enterprise is custom. Extra credits are bought at the underlying API rate with zero markup. |
| DeepSeek | 自有模型，但不锁定Built around DeepSeek models (deepseek-flash and deepseek-v4-pro on DeepSeek's API). Also third-party providers such as Anthropic, OpenAI, Kimi and GLM, and any custom OpenAI-compatible or Anthropic-style endpoint | 开源 | Desktop app (macOS download on the vendor page), Web UI, started with one npx command, Terminal (CLI modes) | The harness is MIT-licensed and free; you pay your model provider. On DeepSeek's own API, off-peak rates are half the peak rates. |
| Earendil (created by Mario Zechner) | 任意模型Many providers (Anthropic, OpenAI, Google, Azure, Bedrock, Mistral, Groq, xAI, OpenRouter, Ollama and more) through API keys, or by logging in with an existing subscription | 开源 | Terminal, Print and JSON mode, RPC, TypeScript SDK | Free and MIT-licensed; you pay your model provider or use a subscription you already have. A paid Pi plan: NOT PUBLISHED. |
| Cognition | 自有模型，但不锁定SWE-2 (Cognition's own model, free to use on Pro) plus frontier models from OpenAI, Claude, Gemini and SpaceXAI and leading open-source models; paid plans can also use your ChatGPT plan | 专有 | Devin Cloud (autonomous cloud agent), Devin Desktop, CLI, Slack, Teams, Linear, Jira and Git hosts | Free $0 (light quota); Pro $20/month; Max $200/month; Teams $80/month plus $40/month per full developer seat; Enterprise on request. Extra usage is bought at API pricing. |
| 自有模型，但不锁定Gemini models | 专有 | Desktop, CLI, SDK | Via Google AI Pro / Ultra and Gemini Code Assist licences |
Oh My Pi (omp)
can1357 (community)- Hashline edits cut output tokens
- The model patches by content-hash anchor instead of retyping lines, and a stale anchor is rejected before it corrupts the file. The project reports Grok 4 Fast spending 61% fewer output tokens on the same work (self-reported) 来源
- Real IDE machinery, not a text box
- 31 built-in tools, 14 LSP operations and 28 DAP (debugger) operations, so renames go through the language server and it can drive lldb, dlv or debugpy 来源
- Adopts your existing config on first run
- Inherits rules, skills and MCP servers already on disk from .claude, .cursor, .codex, .gemini, .cline, .windsurf and .github/copilot — no migration script 来源
扩展点
- 支持技能
- 支持钩子
- 支持子智能体
- 支持插件
- 支持MCP
真实的限制
Community-run and moving very fast — three releases shipped on 1 Sep 2026 alone — with no vendor support contract, no SLA and no company behind it. It is also the most complex harness here: ten model roles and 31 tools are a lot of surface to learn, and the headline efficiency numbers are the project's own, not independently verified.
最适合
Technical teams who want IDE-grade tooling and a hard floor on spend, and who can absorb a fast-moving open-source dependency without vendor support.
来源 已验证 2026-09-02
人们真正会问的问题
- 什么是 AI 编程框架？
- 将语言模型变成智能体的客户端。它运行循环，提供文件编辑和终端等工具，管理上下文和权限，并决定任务何时完成。模型提供推理；框架提供其他一切。
- 框架会改变我的付费金额吗？
- 会，而且影响很大。框架控制它重新发送多少上下文、是否缓存，以及需要多少轮次。两个框架在相同任务上运行同一模型，输出 token 可能相差数倍，而输出 token 占账单的大部分。
- 我可以在一个框架中使用多个模型提供商吗？
- 可以，前提是选择不绑定特定模型的工具。OpenCode、Pi 和 Command Code 支持你使用不同服务商的自有密钥；DeepSeek Harness 支持第三方服务商和自定义端点。Claude Code 仅支持 Anthropic 模型；ZCode 和 Antigravity 围绕各自供应商的模型构建；Devin 则在自己的方案中提供一组选项。
- Gemini CLI 还能用吗？
- 消费者套餐不可用。它已于 2026 年 6 月 18 日停止为 Google AI Pro 和 Ultra 订阅者以及免费 Gemini Code Assist 用户处理请求，并由 Antigravity CLI 取代。Code Assist 标准版或企业版许可证下的访问不受影响。
框架选择是模型选择的下游。
选择能以最低成本达到您质量标准的模型，然后选择以您团队工作方式运行它的框架。反过来做，就会让人最终锁定到他们本无意选择的供应商。
最近发布了什么
来自每家主要实验室的带日期发布，并将各供应商自己的说法标注为说法。
打开独立评分
第三方基准测试结果——刻意与实验室自己发布的任何内容分开。
打开实际要构建什么
部署计划会按您的预算为一个工作流挑选技术栈，包括模型和工具。
打开我们为本页上的每项声明记录来源和核查日期，并展示分歧，而不是悄悄解决它们。 我们如何验证