SmolLM2 360M

SmolLM2 360M loads in 0.5 GB at Q4_K_M and is comfortable from 8 GB of memory, Macs included from the Mac Mini M6 16GB. Below: the full memory math, the cheapest card that runs it well, and the fastest machine we track.

PARAMETERS
0.36B
FORMAT
Q4_K_M
MIN MEMORY
1 GB
BEST FOR
Chat, Embedded

Memory math

Weights (Q4_K_M)
0.5 GB
+ KV cache (16k)
~1.8 GB
Total at 16k
~2.3 GB
Comfortable from
8 GB

KV = fp16 estimate (q8_0 cache roughly halves it). "Comfortable" = weights + KV within the engine's tiered budget (~70-85% of memory).

Hardware snapshot

Cheapest GPU
~300 tok/s est. — ~$550 used
Fastest
~300 tok/s est.
Runs on a Mac?
~180 tok/s est.

Frequently asked questions

How much memory does SmolLM2 360M need?

0.5 GB for the Q4_K_M weights, plus ~1.8 GB of KV-cache at 16k context — about 2.3 GB total. Comfortable from 8 GB of VRAM or unified memory.

Does SmolLM2 360M run on a Mac?

Yes — from the Mac Mini M6 16GB (~180 tok/s est.). Unified memory means the RAM budget is the only limit.

What is the cheapest GPU for SmolLM2 360M?

The AMD Radeon RX 7900 XT is the cheapest tracked card that runs SmolLM2 360M comfortably — ~300 tok/s est. at ~$550 used.

Cite this page

ModelFit: SmolLM2 360M — specs, memory math and hardware verdicts.
https://modelfit.io/models/smollm2-360m/ (dataset updated 2026-09-03, CC BY 4.0).