🏠
Developer
Local Execution
Can be downloaded and run locally on your own hardware without relying on cloud APIs or paying token fees.
Which AI has Local Execution?
17 of 108 tools in the CanIUse matrix support Local Execution. Open a tool-by-feature answer for the exact plan and source.
Which AI tools support this?
| Tool | Status | Notes | Link |
|---|---|---|---|
| 🤖 ChatGPT | No | Closed weights; runs only on OpenAI infrastructure. | Try ChatGPT (opens in a new tab) |
| 🧠 Claude | No | Closed weights; runs only on Anthropic infrastructure and partner clouds. | Try Claude (opens in a new tab) |
| ✨ Google Gemini | Partial | Gemini itself is closed, but the smaller Gemma models are downloadable and run locally. | Try Google Gemini (opens in a new tab) |
| 🐋 DeepSeek | Yes | Open weights can be self-hosted; the largest variants need substantial GPU memory. | Try DeepSeek (opens in a new tab) |
| 🤖 Microsoft Copilot | No | Microsoft Copilot runs in the cloud, not as a local model you download. | Try Microsoft Copilot (opens in a new tab) |
| 🌬️ Mistral AI | Partial | Several models ship open weights; the largest ones are API-only. | Try Mistral AI (opens in a new tab) |
| 🦙 Llama | Yes | Weights are downloadable and run locally via llama.cpp, Ollama and similar runtimes. | Try Llama (opens in a new tab) |
| 🎯 Cohere | Partial | Command models publish open weights for research; commercial use runs through the API or a private deployment. | Try Cohere (opens in a new tab) |
| 🤗 HuggingChat | Yes | Serves open models that can equally be downloaded and run locally. | Try HuggingChat (opens in a new tab) |
| ⌨️ Claude Code | No | The CLI runs on your machine; the model stays on Anthropic (or a third-party provider) infrastructure. | Try Claude Code (opens in a new tab) |
| ☄️ Comet | No | The browser runs locally; model inference stays on Perplexity’s cloud. | Try Comet (opens in a new tab) |
| 🎞️ Veo 3 | No | Hosted by Google. Weights are not for local download. | Try Veo 3 (opens in a new tab) |
| 🎥 Kling AI | No | Hosted by Kuaishou. Weights are not for local download. | Try Kling AI (opens in a new tab) |
| 📓 NotebookLM | No | The new “cloud computer” is hosted by Google, not a local runtime. | Try NotebookLM (opens in a new tab) |
| 🛸 Google Antigravity | Yes, requires free plan | Agents run locally on the machine; models remain Google-hosted Gemini (and partners). | Try Google Antigravity (opens in a new tab) |
| 🧭 Kiro | Yes, requires free plan | IDE and Crew run on your machine; web sandboxes are cloud. Models are hosted. | Try Kiro (opens in a new tab) |
| ⬛ FLUX | Yes, requires free plan | Official GitHub inference repo runs open-weight FLUX.1 [schnell] and [dev] on your hardware. | Try FLUX (opens in a new tab) |
| 🌀 Stable Diffusion | Yes, requires free plan | Weights and inference code are on Hugging Face/GitHub. SD 3.5 Medium is documented at ~9.9 GB VRAM. | Try Stable Diffusion (opens in a new tab) |
| 🔤 Ideogram | Partial | Developer overview lists open weights for Ideogram 4.0. Hosted API remains the documented production path. | Try Ideogram (opens in a new tab) |
| 🐰 CodeRabbit | Partial | CLI/IDE review local diffs; Enterprise offers self-hosting. Models are otherwise hosted. | Try CodeRabbit (opens in a new tab) |
| 🧃 Replit | No | Code runs on Replit’s cloud, not as a local runtime you download. | Try Replit (opens in a new tab) |
| 🤝 Aider | Yes, requires free plan | The program runs on your machine. Models can be local or cloud APIs you configure. | Try Aider (opens in a new tab) |
| 🚀 Warp | Yes, requires free plan | The terminal and Agent CLI run locally. Cloud agents are optional. Homepage also says the terminal is open source. | Try Warp (opens in a new tab) |
| 🔷 Tabnine | Yes, requires plus plan | On-prem/air-gapped deploy plus local IDE/CLI. Unlimited usage when using your own LLM on-prem. | Try Tabnine (opens in a new tab) |
| 📝 Grammarly | No | Cloud service. Enterprise BYOK is key management, not a local model. | Try Grammarly (opens in a new tab) |
| 🔗 n8n | Yes, requires free plan | Community Edition self-hosts (Docker). Cloud is optional. | Try n8n (opens in a new tab) |
| ⚡ Groq | No | Runs on Groq LPUs in their cloud, not on your GPU. | Try Groq (opens in a new tab) |
| 🔀 OpenRouter | No | Routes to cloud providers. Not a local runtime. | Try OpenRouter (opens in a new tab) |
| 🤗 Hugging Face | Yes, requires free plan | Transformers/Diffusers/llama.cpp-style downloads. Transformers.js runs in the browser. | Try Hugging Face (opens in a new tab) |
| 🎙️ Whisper | Yes, requires free plan | Official GitHub repo runs on your hardware. | Try Whisper (opens in a new tab) |
| 🦞 OpenClaw | Yes, requires free plan | Open-source agent that runs on your machines. Cloud models are optional BYOK. | Try OpenClaw (opens in a new tab) |
| 🦦 Otter.ai | No | Cloud SaaS. Desktop recorder still sends audio to Otter. | Try Otter.ai (opens in a new tab) |
| ⚡ Zapier | No | Cloud SaaS. Not self-hosted. | Try Zapier (opens in a new tab) |
| ✨ Genspark | No | Hosted web workspace, not a local model runtime. | Try Genspark (opens in a new tab) |
| 🔥 Fireflies.ai | No | Cloud SaaS. | Try Fireflies.ai (opens in a new tab) |
| 🌊 Fathom | No | Cloud SaaS. | Try Fathom (opens in a new tab) |
| 🖼️ Photoroom | No | Cloud editor + API. | Try Photoroom (opens in a new tab) |
| 🦜 LangChain | Yes, requires free plan | Open-source frameworks run locally. LangSmith cloud is optional. | Try LangChain (opens in a new tab) |
| ♻️ Replicate | Partial | Cog is open source so you can run locally. Hosted Replicate is cloud GPUs. | Try Replicate (opens in a new tab) |
| 🤝 Together AI | No | Cloud GPUs / serverless. Not on your laptop. | Try Together AI (opens in a new tab) |
| 📐 Linear | No | Cloud SaaS. Agent coding sessions run in Linear’s environment. | Try Linear (opens in a new tab) |
| ✅ ClickUp | No | Cloud SaaS. | Try ClickUp (opens in a new tab) |
| 🌸 Asana | No | Cloud SaaS. | Try Asana (opens in a new tab) |
| 🔗 Make | No | Cloud SaaS, unlike self-hosted n8n. | Try Make (opens in a new tab) |
| 🎨 Figma | No | Cloud editor. Make-on-local-codebase is still a Figma feature, not self-hosted Figma. | Try Figma (opens in a new tab) |
| ✨ Adobe Express | No | Cloud editor. | Try Adobe Express (opens in a new tab) |
| 🪶 QuillBot | No | Cloud web app. | Try QuillBot (opens in a new tab) |
| 📝 Wordtune | No | Cloud + extension. | Try Wordtune (opens in a new tab) |
| 📊 Beautiful.ai | No | Cloud SaaS. | Try Beautiful.ai (opens in a new tab) |
| 📕 Tome | No | Nothing to run. URL 404. | Try Tome (opens in a new tab) |
| 🪨 Amazon Bedrock | No | Managed AWS inference. Distinct from Kiro local dev. | Try Amazon Bedrock (opens in a new tab) |
| ☁️ Azure OpenAI Service | No | Cloud-hosted in your Azure region. | Try Azure OpenAI Service (opens in a new tab) |
| 🔺 Vertex AI | No | Cloud GCP inference. Distinct from Antigravity local dev. | Try Vertex AI (opens in a new tab) |
| 🌲 Pinecone | No | Managed cloud vector DB. | Try Pinecone (opens in a new tab) |
| 🗣️ D-ID | No | Cloud-rendered avatar video. | Try D-ID (opens in a new tab) |
| 🎙️ Deepgram | Yes, requires enterprise plan | Self-hosted deployment option alongside cloud. | Try Deepgram (opens in a new tab) |
| 🎧 AssemblyAI | No | Cloud-only STT. Distinct from self-hosted Whisper. | Try AssemblyAI (opens in a new tab) |
| 📻 Adobe Podcast | No | Cloud web tool. | Try Adobe Podcast (opens in a new tab) |
| 🎵 Stable Audio | Partial | Community ONNX CPU inference mentioned; hosted API is cloud. | Try Stable Audio (opens in a new tab) |
| 🎬 InVideo | No | Cloud-rendered video editor. | Try InVideo (opens in a new tab) |
| ⌨️ Raycast AI | Yes, requires pro plan | Ollama local models run on-device. | Try Raycast AI (opens in a new tab) |
| ⚡ Superhuman | No | Cloud-synced mail and docs. | Try Superhuman (opens in a new tab) |