// SmolLM / SmolVLM · Text LLM
SmolLM2-1.7B: VRAM & local deployment
1.71B parameters. About 3.3 GB at 4-bit and 5.9 GB at BF16 with one 8K-token request.
released 2024-10-30licence: Commercial OKmemory figures are estimates
01Specs
| Publisher | Hugging Face · HuggingFaceTB |
|---|---|
| Family / series | SmolLM / SmolVLM · SmolLM2 |
| Type | Text LLM (text-generation) |
| Parameters | 1.71B Source: counted from official safetensors weights |
| Context length | 8K tokens (8,192)max_position_embeddings in config.json |
| Architecture | 24 layers · GQA |
| Released weights | BF16 · 3.4 GBdownload |
| Licence | Apache-2.0 · Commercial OKPermissive licence (Apache-2.0 / MIT / CC-BY / OpenMDW): commercial use allowed. |
| Released | 2024-10-30 (Hugging Face repo created) |
02How much VRAM it needs (estimate)
Total = weights + 10% + 0.5GB overhead + KV cache for one request (FP16 KV). More concurrent users multiply the KV column.
| Precision | Weights | Total @ 8K | Smallest GPU | Mac |
|---|---|---|---|---|
| BF16 / FP16 (original) | 3.4 GB | 5.9 GB | 8GB GPU | Mac 16GB |
| FP8 / INT8 / Q8_0 | 1.8 GB | 4.1 GB | 8GB GPU | Mac 16GB |
| 4-bit (Q4_K_M / AWQ / GPTQ-Int4) | 1.0 GB | 3.3 GB | 8GB GPU | Mac 16GB |
KV cache per request: 8K 1.6 GB
Bits per weight: BF16 16, FP8 8.5, 4-bit 4.85. “Fits” = total ≤ 90% of memory; Mac GPU can use ~2/3 of unified memory (≤36GB) or ~3/4 (larger). For MoE, all experts must be in memory, so size by total parameters. Method →
03How to run it
Only runtimes the publisher or the runtime’s official library names for this model.
- Transformers — mentioned in the official model card
- llama.cpp — official GGUF weights
- MLX — official MLX weights
- Ollama — in the official Ollama library · smollm2:1.7b
ollama run smollm2:1.7b04Official repositories & quantised builds
5 official repo(s) from HuggingFaceTB.
- SmolLM2-1.7B-Instruct Instruct
- SmolLM2-1.7B main
- SmolVLM-Instruct Instruct
- SmolLM2-1.7B-Instruct-GGUF Instruct/GGUF
- SmolLM2-1.7B-Instruct-Q8-mlx Instruct/MLX
05Other SmolLM / SmolVLM models
All 8 →| Model | Params | 4-bit total | Released |
|---|---|---|---|
| SmolLM2-135M · same series | 135M | 0.8 GB | 2024-10-31 |
| SmolLM2-360M · same series | 362M | 1.1 GB | 2024-10-31 |
| SmolLM3-3B | 3.08B | 3.2 GB | 2025-07-08 |
| SmolVLM2-2.2B-Instruct | 2.25B | 2.0 GB | 2025-02-08 |
| SmolLM-135M | 135M | 0.8 GB | 2024-07-14 |
| SmolLM-360M | 362M | 1.1 GB | 2024-07-14 |
| SmolLM-1.7B | 1.71B | 3.3 GB | 2024-07-14 |
06Sources
Data checked 2025-02-06. Parameters, context and licence are copied from the official repository; memory figures are our estimates.