Which models run on your PC
Grouped by graphics card, so you can find yours. Bobbin checks your actual hardware when you install it and tells you which of these will run well, which will run slowly, and which will not run at all.
8GB graphics card
RTX 3060 Ti, RTX 4060, RTX 2070The most common tier. Everything here leaves room for a long conversation.
Qwen3 8Bqwen
The best all-rounder for an 8GB card
Released April 2025 · 1 year 5 months oldApache-2.012.7M downloads/month
5.0 GB download
needs 5.4 GB(estimated)
Qwen2.5 7B Instructqwen
The previous generation's all-rounder
Released September 2024 · 2 years oldApache-2.09.6M downloads/month
4.7 GB download
needs 5.3 GB
Qwen2.5 0.5B Instructqwen
Tiny and instant — runs on laptops without a graphics card
Released September 2024 · 2 years oldApache-2.08.6M downloads/month
398 MB download
needs 700 MB
Llama 3.1 8B Instructmeta
Meta's workhorse — the base most add-ons target
Released July 2024 · 2 years 2 months oldLlama 3.1 Community License6.1M downloads/month
4.9 GB download
needs 6.1 GB
Gemma 4 E4Bgoogle
Google's newest small model
Released March 2026 · 6 months oldApache-2.04.5M downloads/month
5.0 GB download
needs 5.3 GB(estimated)
Qwen2.5 Coder 7B Instructqwen
Code completion and review, locally
Released September 2024 · 2 years oldApache-2.02.5M downloads/month
4.7 GB download
needs 5.3 GB
Llama 3.2 3B Instructmeta
Small, fast, runs on almost anything
Released September 2024 · 2 years oldLlama 3.2 Community License1.9M downloads/month
2.0 GB download
needs 3.3 GB
Gemma 3 4Bgoogle
Tiny and fast — runs on laptops without a graphics card
Released February 2025 · 1 year 7 months oldGemma Terms of Use1.7M downloads/month
2.5 GB download
needs 3.0 GB(estimated)
Phi-4 Minimicrosoft
Small and quick, punches above its weight on reasoning
Released February 2025 · 1 year 7 months oldMIT378k downloads/month
2.5 GB download
needs 3.0 GB(estimated)
DeepSeek R1 Distill 7Bdeepseek
Thinks step by step before answering
Released January 2025 · 1 year 8 months oldMIT245k downloads/month
4.7 GB download
needs 5.0 GB(estimated)
Ministral 3 8Bmistral
European-built, strong on French and German
Released October 2025 · 11 months oldApache-2.075k downloads/month
5.2 GB download
needs 5.5 GB(estimated)
16GB graphics card
RTX 4060 Ti 16GB, RTX 5060 Ti, RX 7800 XTNoticeably more capable models, and enough room to work through long documents.
Qwen3 14Bqwen
The all-rounder, with more depth
Released April 2025 · 1 year 5 months oldApache-2.02.1M downloads/month
9.0 GB download
needs 9.1 GB(estimated)
Phi-4 14Bmicrosoft
Trained for reasoning rather than trivia
Released December 2024 · 1 year 9 months oldMIT627k downloads/month
9.1 GB download
needs 9.1 GB(estimated)
Gemma 3 12Bgoogle
Careful, readable writing
Released March 2025 · 1 year 6 months oldGemma Terms of Use601k downloads/month
7.3 GB download
needs 7.5 GB(estimated)
DeepSeek R1 Distill 14Bdeepseek
Serious reasoning, if you can wait for it
Released January 2025 · 1 year 8 months oldMIT348k downloads/month
9.0 GB download
needs 9.0 GB(estimated)
Mistral Small 3.2 24Bmistral
The most capable model that still fits 16GB
Released June 2025 · 1 year 3 months oldApache-2.0196k downloads/month
14.3 GB download
needs 14.0 GB(estimated)
Mellum2 12Bjetbrains
Built by JetBrains, for code
Released May 2026 · 4 months oldApache-2.03k downloads/month
8.1 GB download
needs 8.2 GB(estimated)
24GB graphics card
RTX 3090, RTX 4090, RTX 5090The best you can run on a single consumer card.
Gemma 4 26B A4Bgoogle
26B of knowledge, a fraction of the work per word
Released March 2026 · 6 months oldApache-2.010.3M downloads/month
16.9 GB download
needs 16.5 GB(estimated)
Gemma 4 31Bgoogle
Google's flagship open model
Released March 2026 · 6 months oldApache-2.09.2M downloads/month
18.3 GB download
needs 17.7 GB(estimated)
Qwen3.8 27Bqwen
The strongest all-rounder you can run on one card
Released August 2026 · 1 month oldApache-2.07.1M downloads/month
16.5 GB download
needs 16.0 GB(estimated)
Qwen3 32Bqwen
The largest dense model that fits 24GB
Released April 2025 · 1 year 5 months oldApache-2.05.0M downloads/month
19.8 GB download
needs 19.0 GB(estimated)
Qwen3 30B A3Bqwen
30B of knowledge, 3B of work per word
Released April 2025 · 1 year 5 months oldApache-2.01.7M downloads/month
18.6 GB download
needs 18.0 GB(estimated)
Gemma 3 27Bgoogle
The previous Gemma flagship, widely tested
Released March 2025 · 1 year 6 months oldGemma Terms of Use393k downloads/month
16.5 GB download
needs 16.1 GB(estimated)
Sizes are for the recommended version of each model and are read from the file itself. Memory requirements marked as estimates have not yet been run on that class of card — the app says so rather than presenting a guess as a measurement. Download figures come from Hugging Face and cover the last 30 days.