PowerMoE-3b vs SmolLM-1.7B-Instruct-quantized.w4a16
Specs, VRAM requirements and download trends — updated 8 August 2026.
ibm-research · A · Open weights
Params
3.4 B
Context
4,096
30d
171 K
apache-2.0 · commercial OK
nm-testing · B · Open weights
Params
1.8 B
Context
2,048
30d
152 K
apache-2.0 · commercial OK
Specification comparison
Differences are highlighted; identical values are muted.
| Specification | PowerMoE-3b | SmolLM-1.7B-Instruct-quantized.w4a16 |
|---|---|---|
| Parameters | 3.4 B | 1.8 B |
| Architecture | GraniteMoeForCausalLM | LlamaForCausalLM |
| Context length | 4,096 | 2,048 |
| Licence | apache-2.0 | apache-2.0 |
| Commercial use | Allowed | Allowed |
| Languages | — | 1 |
| Downloads 30d | 170,877 | 151,608 |
| Downloads all time | 12.0 M | 5.8 M |
| Quantizations on the Hub | SAFETENSORS | SAFETENSORS |
| Gated | No | No |
| First seen on the Hub | 2024-08-14 | 2024-08-23 |
Download trend
Daily snapshots, last 90 days (28 Jul – 25 Oct)
1.8 M152 K
PowerMoE-3b
SmolLM-1.7B-Instruct-quantized.w4a16
A overtook B on 24 Oct
VRAM side by side
On RTX 4090 · 24 GB · 8K context unless noted
SmolLM-1.7B-Instruct-quantized.w4a16 · fp16
2.7 GB / 24 GB
PowerMoE-3b · fp16
15.9 GB / 24 GB