DeepSeek-V4-Flash-NVFP4 vs Llama-3.1-405B-FP8
Specs, VRAM requirements and download trends — updated 6 August 2026.
nvidia · A · Open weights
Params
166.7 B
Context
1,048,576
30d
203 K
mit · commercial OK
meta-llama · B · Open weights
Params
405.9 B
Context
—
30d
465 K
llama3.1 licence · conditional
Specification comparison
Differences are highlighted; identical values are muted.
| Specification | DeepSeek-V4-Flash-NVFP4 | Llama-3.1-405B-FP8 |
|---|---|---|
| Parameters | 166.7 B | 405.9 B |
| Architecture | DeepseekV4ForCausalLM | LlamaForCausalLM |
| Context length | 1,048,576 | — |
| Licence | mit | llama3.1 |
| Commercial use | Allowed | Conditional |
| Languages | — | 8 |
| Downloads 30d | 202,991 | 465,185 |
| Downloads all time | 1.9 M | 2.7 M |
| Quantizations on the Hub | SAFETENSORS | SAFETENSORS |
| Gated | No | Yes — accept licence |
| First seen on the Hub | 2026-05-18 | 2024-07-20 |
Download trend
Daily snapshots, last 90 days (13 Aug – 10 Nov)
772 K203 K
DeepSeek-V4-Flash-NVFP4
Llama-3.1-405B-FP8
VRAM side by side
On RTX 4090 · 24 GB · 8K context unless noted
Llama-3.1-405B-FP8 · fp16
597.3 GB / 24 GB
DeepSeek-V4-Flash-NVFP4 · fp16 @ 1024K ctx
3,386.8 GB / 24 GB