OLMo 2 32B
Earlier fully open model. Brilliant and honest, but forgets what you said a few pages back.
Every model below runs on your own computer. Click a name for its license, sources, and setup steps.
| Model | Maker | Trust | Quality | RAM needed | Business | Sign-up |
|---|---|---|---|---|---|---|
| Olmo 3 7B | Ai2 | 100 | Good | 8 GB | Yes | No |
| Olmo 3.1 32B | Ai2 | 100 | Very strong | 26 GB | Yes | No |
| OLMo 2 32B | Ai2 | 100 | Very strong | 26 GB | Yes | No |
| SmolLM3 3B | Hugging Face | 100 | Basic | 5 GB | Yes | No |
| GPT-OSS 20B | OpenAI | 84 | Very strong | 17 GB | Yes | No |
| GPT-OSS 120B | OpenAI | 84 | Top home tier | 87 GB | Yes | No |
| Qwen3.5 9B | Alibaba Qwen | 84 | Good | 9 GB | Yes | No |
| Qwen3.8 27B | Alibaba Qwen | 84 | Very strong | 23 GB | Yes | No |
| Qwen3 4B 2507 | Alibaba Qwen | 84 | Good | 5 GB | Yes | No |
| Qwen3 8B | Alibaba Qwen | 84 | Good | 8 GB | Yes | No |
| Qwen3 14B | Alibaba Qwen | 84 | Strong | 13 GB | Yes | No |
| Qwen3 30B-A3B 2507 | Alibaba Qwen | 84 | Very strong | 24 GB | Yes | No |
| Qwen3 32B | Alibaba Qwen | 84 | Very strong | 26 GB | Yes | No |
| Qwen3 Coder 30B | Alibaba Qwen | 84 | Very strong | 24 GB | Yes | No |
| Qwen2.5 VL 7B | Alibaba Qwen | 84 | Good | 8 GB | Yes | No |
| Gemma 4 12B | 84 | Strong | 11 GB | Yes | No | |
| Gemma 4 E4B | 84 | Good | 8 GB | Yes | No | |
| DeepSeek R1 Distill 7B | DeepSeek | 84 | Good | 8 GB | Yes | No |
| DeepSeek R1 Distill 14B | DeepSeek | 84 | Strong | 13 GB | Yes | No |
| DeepSeek R1 0528 8B | DeepSeek | 84 | Good | 8 GB | Yes | No |
| Phi-4 | Microsoft | 84 | Strong | 13 GB | Yes | No |
| Phi-4 mini | Microsoft | 84 | Basic | 5 GB | Yes | No |
| Mistral 7B v0.3 | Mistral AI | 84 | Good | 8 GB | Yes | No |
| Mistral Nemo 12B | Mistral AI | 84 | Strong | 11 GB | Yes | No |
| Mistral Small 3.2 | Mistral AI | 84 | Very strong | 20 GB | Yes | No |
| Magistral Small | Mistral AI | 84 | Very strong | 19 GB | Yes | No |
| Devstral Small | Mistral AI | 84 | Very strong | 19 GB | Yes | No |
| Granite 3.3 2B | IBM | 84 | Basic | 4 GB | Yes | No |
| Granite 3.3 8B | IBM | 84 | Good | 8 GB | Yes | No |
| Granite 4.0 H Tiny | IBM | 84 | Good | 8 GB | Yes | No |
| Granite 4.0 H Small | IBM | 84 | Very strong | 26 GB | Yes | No |
| GLM-4.5 Air | Z.ai | 84 | Top home tier | 82 GB | Yes | No |
| Hermes 3 8B | Nous Research | 64 | Good | 8 GB | With limits | No |
| Nemotron Nano 9B v2 | NVIDIA | 64 | Good | 9 GB | With limits | No |
| LFM2 1.2B | Liquid AI | 64 | Basic | 3 GB | With limits | No |
| Llama 3.1 8B | Meta | 58 | Good | 8 GB | With limits | Yes |
| Llama 3.2 3B | Meta | 58 | Basic | 5 GB | With limits | Yes |
| Llama 3.3 70B | Meta | 58 | Top home tier | 53 GB | With limits | Yes |
| Llama 4 Scout | Meta | 58 | Top home tier | 81 GB | With limits | Yes |
| Gemma 3 4B | 58 | Good | 6 GB | With limits | Yes | |
| Gemma 3 12B | 58 | Strong | 11 GB | With limits | Yes | |
| Gemma 3 27B | 58 | Very strong | 22 GB | With limits | Yes | |
| Gemma 3n E4B | 58 | Good | 8 GB | With limits | Yes | |
| Command R7B | Cohere | 38 | Good | 8 GB | No | Yes |
Trust and quality use our published method. RAM is an estimate for a 4-bit file.
By maker
Ai2
Olmo 3 7B
Fully open chat model. Data and code are public. Fits on the laptop you already own, if it has 8 GB.
Olmo 3.1 32B
Largest fully open model here. Data and code are public. Nothing up its sleeve.
Alibaba Qwen
Qwen2.5 VL 7B
Show it a screenshot and it will tell you what is going on. Fits in 8 GB.
Qwen3 14B
The 8B's older sibling. Smarter, and wants about 13 GB.
Qwen3 30B-A3B 2507
Big on paper, quick in practice. Only part of it works on each word.
Qwen3 32B
A careful thinker. Wants about 26 GB and a little patience.
Qwen3 4B 2507
Tiny, quick, and fits in 8 GB. The pocket knife of this list.
Qwen3 8B
Thinks step by step on a normal laptop. Fits in 8 GB.
Qwen3 Coder 30B
Writes code so you do not have to. You still have to read it.
Qwen3.5 9B
Keeps a lot of your chat in mind at once. Fits a 16 GB machine.
Qwen3.8 27B
The all-rounder for a 32 GB machine. Does a bit of everything, complains about nothing.
Cohere
Command R7B
Cites its sources like a good student. Not for business use, sadly.
DeepSeek
DeepSeek R1 0528 8B
A newer small thinker. MIT license, so legal can relax.
DeepSeek R1 Distill 14B
Thinks out loud before it answers. Get comfy.
DeepSeek R1 Distill 7B
Shows its work, like your math teacher always wanted. Fits in 8 GB.
Gemma 3 12B
A good writer that can also look at pictures. Wants about 11 GB.
Gemma 3 27B
Strong writer for a 32 GB machine. Google's rules still apply.
Gemma 3 4B
Small, reads images, and fits in 8 GB. Google's rules apply.
Gemma 3n E4B
Built for phones. Next to it, your laptop will feel like a supercomputer.
Gemma 4 12B
Google's open model with an Apache license. Fits a 16 GB machine.
Gemma 4 E4B
Small and efficient. Your laptop will barely notice it is there.
Hugging Face
SmolLM3 3B
Small, fully open, training data mix is public. Also has the cutest name on this list.
IBM
Granite 3.3 2B
Tiny business helper. About 4 GB, which is basically nothing.
Granite 3.3 8B
IBM's business model. It wears a tie to the chat.
Granite 4.0 H Small
Bigger IBM model for long documents. Wants about 26 GB.
Granite 4.0 H Tiny
Light on memory, even with long documents. About 8 GB.
Liquid AI
LFM2 1.2B
Very small. Runs on the laptop you were about to recycle.
Meta
Llama 3.1 8B
The one everybody has heard of. Free to use, but Meta writes the rules, so read them.
Llama 3.2 3B
Small enough for a modest laptop. Big enough to write your out-of-office reply.
Llama 3.3 70B
Smart, but wants about 53 GB of memory. Not a laptop guest.
Llama 4 Scout
Reads images too. Wants about 81 GB, so bring a big machine and snacks.
Microsoft
Phi-4
Good at math, short memory. Keep your chats short and sweet.
Phi-4 mini
Small model, long memory window. Fits in 8 GB.
Mistral AI
Devstral Small
Fixes bugs across many files. Your code reviewer will still have notes.
Magistral Small
Mistral's step-by-step thinker. Apache license, no drama.
Mistral 7B v0.3
An oldie but a goodie. Light, well known, Apache license.
Mistral Nemo 12B
Speaks many languages. Fits a 16 GB machine.
Mistral Small 3.2
Chats and reads images. Wants about 20 GB.
NVIDIA
Nemotron Nano 9B v2
You can switch its thinking on or off. If only people came with that button.
Nous Research
Hermes 3 8B
Built on Llama 3.1, so Meta's rules come along for the ride.
OpenAI
GPT-OSS 120B
Hard reasoning and agent tasks on a high-end workstation. Plan on a 128 GB machine. Your laptop just flinched.
GPT-OSS 20B
OpenAI, but you get to keep it at home. Plan on about 17 GB of memory.
Z.ai
GLM-4.5 Air
Strong agent and coding work. It is called Air and it wants a 128 GB machine. Someone in naming has jokes.