Open-source and local AI tools
Run AI on your own computer: private, offline, and free apart from your hardware. 11 of our 72 tools are open source, have open model weights, or run models locally.
Quick start
- Easiest: install LM Studio or Jan, pick a model, and chat. No account needed.
- For developers: install Ollama and run a model from the command line or its API; add Open WebUI for a browser interface your team can share.
- Images: ComfyUI with the FLUX.1 [schnell] model. Transcription: Whisper.
Want three local picks for your exact project?
Use the matcher (open source & local only)AI assistants
| Tool | Open source / local | Price | Notes | Checked | |
|---|---|---|---|---|---|
| DeepSeek Free reasoning and math chat | Open weights: MIT (DeepSeek-R1) | See vendor pricing | Chat app is cloud; distilled R1 models run locally via Ollama | source 2026-10-23 | Get DeepSeek → |
| Ollama Running open models like Llama, Qwen, Gemma, DeepSeek and gpt-oss on your own computer | Open source: MIT | Free locally (cloud Pro $20/mo) ✓ checked | Command-line and desktop runner for open models; local use is unlimited | source 2026-10-23 | Get Ollama → |
| LM Studio Easiest desktop app to download and chat with local models privately | Runs locally (Proprietary app, free for personal and internal business use) | Free for home and work ✓ checked | Not open source, but everything runs on your device | source 2026-10-23 | Get LM Studio → |
| Jan Open-source, offline ChatGPT-style desktop app | Open source: Apache 2.0 | Free (open source) ✓ checked | Fully offline desktop chat app | source 2026-10-23 | Get Jan → |
| Open WebUI Self-hosted ChatGPT-style interface for a team, on local or cloud models | Open source: Open WebUI License (BSD-3 plus branding clause; not OSI-approved) | Free to self-host ✓ checked | Self-hosted web interface; pairs with Ollama | source 2026-10-23 | Get Open WebUI → |
| Qwen Chat Free chat with Alibaba's open-weight Qwen models | Open weights: Apache 2.0 for most Qwen3 models (some use the Qwen Community License) | Free (API is pay-per-use) ✓ checked | Chat app is cloud; weights run locally via Ollama or LM Studio | source 2026-10-23 | Get Qwen Chat → |
| Mistral Vibe (Le Chat) European AI chat and coding agent from a maker of open-weight models | Open weights: Apache 2.0 for models such as Mistral Small 3.2 | Pro $14.99/mo ✓ checked | Vibe app is cloud; open models run locally via Ollama or LM Studio | source 2026-10-23 | Get Mistral Vibe (Le Chat) → |
Coding & app builders
| Tool | Open source / local | Price | Notes | Checked | |
|---|---|---|---|---|---|
| Aider Open-source AI pair programmer in your terminal, works with local models | Open source: Apache 2.0 | Free (open source; you pay for model API or run local) ✓ checked | Terminal coding assistant; connect a local model via Ollama for fully local use | source 2026-10-23 | Get Aider → |
Image generation
| Tool | Open source / local | Price | Notes | Checked | |
|---|---|---|---|---|---|
| ComfyUI Node-based local image and video generation with open models | Open source: GPL-3.0 | Free (open source) ✓ checked | Runs FLUX, Stable Diffusion and other open image/video models | source 2026-10-23 | Get ComfyUI → |
| FLUX.1 [schnell] Fast open-weight image model you can run locally and use commercially | Open weights: Apache 2.0 (FLUX.1 [dev] is non-commercial) | Free open weights (Apache 2.0) ✓ checked | Run it in ComfyUI | source 2026-10-23 | Get FLUX.1 [schnell] → |
Meetings
| Tool | Open source / local | Price | Notes | Checked | |
|---|---|---|---|---|---|
| Whisper Private, offline speech-to-text and transcription on your own machine | Open source: MIT | Free (open source, MIT) ✓ checked | Speech-to-text code and weights | source 2026-10-23 | Get Whisper → |
Popular open models and their licenses
These are the models you download into Ollama, LM Studio, Jan or ComfyUI.
| Model | License | Notes |
|---|---|---|
| Llama 3.3 / Llama 4 (Meta) | Llama Community License | Free to use; products shipping it must show “Built with Llama”; companies over 700M monthly users need Meta's permission |
| Gemma 4 (Google) | Apache 2.0 | Earlier Gemma versions use the Gemma Terms of Use |
| Qwen3 family (Alibaba) | Apache 2.0 for most models | Some newer variants use the Qwen Community License |
| Mistral Small 3.2 (Mistral AI) | Apache 2.0 | |
| gpt-oss (OpenAI) | Apache 2.0 | |
| DeepSeek-R1 (DeepSeek) | MIT | |
| FLUX.1 [schnell] (Black Forest Labs) | Apache 2.0 | FLUX.1 [dev] is non-commercial |
| Whisper (OpenAI) | MIT | Speech-to-text |
Model licenses checked 2026-10-23 against the model cards, vendor pages and this license summary (checked 2026-10-19). Licenses vary by version, so read the model's own license before commercial use.
“Open source” means the code is published under the license shown; “open weights” means the model files can be downloaded and run on your own hardware (the maker's chat app may still be a cloud service). LM Studio is free and runs models locally but is not open source. Running models locally needs a reasonably modern computer; larger models need a GPU or lots of memory. Licenses were read on the source linked on the check date; check the model's own license before commercial use. Some links are affiliate links (disclosure); they never change what we list.