Loading chapter

MODEL

SIZE

RAM

SPEED

BEST FOR

llama3.2:1b

1.4 GB

2 GB

Very fast

Quick replies, low-end hardware

llama3.2:3bRECOMMENDED

2.0 GB

4–6 GB

Fast

Default for most users

llama3.2-vision:11b

6.5 GB

12 GB

Moderate

img: command, vision tasks

mistral:7b

4.1 GB

8–12 GB

Moderate

Higher-quality writing

codellama:7b

3.8 GB

8–12 GB

Moderate

Code generation and debugging

phi3:14b

8.0 GB

16 GB

Slow

Heavy reasoning, when quality matters most

10 - MODELS

Choosing AImodels.

Pick the brain that fits your hardware and your work. The hero IS the comparison - read across, find your row, download.

MODEL	SIZE	RAM	SPEED	BEST FOR
`llama3.2:1b`	1.4 GB	2 GB	Very fast	Quick replies, low-end hardware
`llama3.2:3b`RECOMMENDED	2.0 GB	4–6 GB	Fast	Default for most users
`llama3.2-vision:11b`	6.5 GB	12 GB	Moderate	img: command, vision tasks
`mistral:7b`	4.1 GB	8–12 GB	Moderate	Higher-quality writing
`codellama:7b`	3.8 GB	8–12 GB	Moderate	Code generation and debugging
`phi3:14b`	8.0 GB	16 GB	Slow	Heavy reasoning, when quality matters most

Disk. Models range from 1 GB to 20 GB+. Smaller is easier to manage and to swap.
RAM. Roughly 2 GB per billion parameters. A 7b model wants 8–12 GB; 70b models want 64 GB+.
Speed vs. quality. Smaller is faster but less sophisticated. Larger is smarter but slower.
Task. Simple tasks → small. Reasoning, code, or vision → larger or specialized.

From the Models page in Typilot, or from the CLI:

ollama pull llama3.2:3b

Once downloaded, return to Typilot. The model appears in the Default Model dropdown under Settings → Connection.

Open Settings.
Pick a different model from Default Model.
Save. Subsequent AI requests use the new model. Already-running requests finish on the old one.