Ollama library

Smallthinker 3B

Ollama library·3B·Transformer·32K ctx·2026-08

Check on my hardwareAdd to compare →Ollama →Hugging Face →

Model size→Hardware requirement→Actual fit

Smallthinker 3B needs ~4.8–8.1 GB of VRAM depending on quantization. The bar is what it needs; the marker is the card's VRAM. Speed figures below are planning estimates.

8 GB

RTX 4060 · 3060 Ti

5.5
8
Q5_K_M · Est. ~20 tok/sComfortable

12 GB

RTX 4070 · 3060 12G

8.1
12
Q8_0 · Est. ~24 tok/sComfortable

16 GB

RTX 4080 · M-series 16G

8.1
16
Q8_0 · Est. ~37 tok/sComfortable

24 GB

RTX 4090 · 3090

8.1
24
Q8_0 · Est. ~56 tok/sComfortable

48 GB

A6000 · 2×24G

8.1
48
Q8_0 · Est. ~56 tok/sComfortable

02  — Quantizations

QuantFile sizeEst. VRAM
Q4_K_M3.6 GB4.8 GB
Q5_K_M4.3 GB5.5 GB
Q8_06.9 GB8.1 GB

03  — Runtimes & use

llama.cpp · Ollama · LM Studio

General · Fast

Memory and speed figures are planning estimates, not measured benchmarks. Memory includes a fixed allowance for KV cache and runtime overhead. Long-context prompts push VRAM use above the numbers shown here.