NVIDIA-Nemotron-3-Nano-4B-BF16 vs Llama-3.2-3B-Instruct-FP8-dynamic

Specs, VRAM requirements and download trends — updated 21 September 2026.

nvidia · A
Params
4.0 B
Context
262,144
30d
3.5 M
RedHatAI · B
Params
3.6 B
Context
131,072
30d
277 K
llama3.2 licence · conditional

Specification comparison

Differences are highlighted; identical values are muted.

Specification NVIDIA-Nemotron-3-Nano-4B-BF16 Llama-3.2-3B-Instruct-FP8-dynamic
Parameters 4.0 B 3.6 B
Architecture NemotronHForCausalLM LlamaForCausalLM
Context length 262,144 131,072
Licence other llama3.2
Commercial use Unknown Conditional
Languages 1 8
Downloads 30d 3,492,652 276,876
Downloads all time 6.9 M 358 K
Quantizations on the Hub SAFETENSORS SAFETENSORS
Gated No No
First seen on the Hub 2026-03-07 2024-09-25

Download trend

Daily snapshots, last 56 days (28 Jul – 21 Sep)

3.5 M207 K
NVIDIA-Nemotron-3-Nano-4B-BF16 Llama-3.2-3B-Instruct-FP8-dynamic

VRAM side by side

On RTX 4090 · 24 GB · 8K context unless noted

Llama-3.2-3B-Instruct-FP8-dynamic · fp16 @ 128K ctx 14.0 GB / 24 GB
✅ Runs comfortably
NVIDIA-Nemotron-3-Nano-4B-BF16 · fp16 @ 256K ctx 28.3 GB / 24 GB
❌ Won’t fit

Adjacent comparisons