Llama-3.1-8B-Instruct vs Gemma-4-26B-A4B-NVFP4
Specs, VRAM requirements and download trends — updated 6 August 2026.
Specification comparison
Differences are highlighted; identical values are muted.
| Specification | Llama-3.1-8B-Instruct | Gemma-4-26B-A4B-NVFP4 |
|---|---|---|
| Parameters | 8.0 B | 14.4 B |
| Architecture | LlamaForCausalLM | Gemma4ForConditionalGeneration |
| Context length | — | — |
| Licence | llama3.1 | apache-2.0 |
| Commercial use | Conditional | Allowed |
| Languages | 8 | — |
| Downloads 30d | 8,014,793 | 1,471,741 |
| Downloads all time | 175.3 M | 5.2 M |
| Quantizations on the Hub | SAFETENSORS | SAFETENSORS |
| Gated | Yes — accept licence | No |
| First seen on the Hub | 2024-07-18 | 2026-05-01 |
Download trend
Daily snapshots, last 10 days (28 Jul – 6 Aug)
8.0 M1.2 M
Llama-3.1-8B-Instruct
Gemma-4-26B-A4B-NVFP4
VRAM side by side
On RTX 4090 · 24 GB · 8K context unless noted
Llama-3.1-8B-Instruct · fp16
19.4 GB / 24 GB
Gemma-4-26B-A4B-NVFP4 · fp16
23.3 GB / 24 GB