SmolLM2-1.7B-Instruct VRAM & Memory Requirements
Specific checkpoints with sourced architecture and weight files. GGUF Q4_K_M (Hugging Face): 2.61 GiB–3.11 GiB at 8,192 tokens, 1 sequence, FP16 KV, llama.cpp.
This result is an estimate, not a measured trace.Not verifiedFile-backed weights
Publisher and provenance
- PublisherHugging Face
- Hugging Face
HuggingFaceTB/SmolLM2-1.7B-Instruct - Revision
31b70e2e869a7173562077fd711b654946d38674 - Total parameters1,711,376,384
- Layers24
- Query heads32
- KV heads32 (mha)
- Head dimension64
- Recorded context capability8,192
- Completenesspartial · full
Memory layers in this estimate
Artifact bytes, raw KV, engine extras, and total demand are separate. This is not a runtime trace.
- Weight artifact
- 0.98 GiB (1,055,609,536 bytes)
- Raw KV cache
- 1.50 GiB (1,610,612,736 bytes)
- Engine extras / workspace
- 0.13 GiB–0.62 GiB
- Total estimated demand
- 2.61 GiB–3.11 GiB
Published artifacts
| Quantization | Selected files | Bytes | Engine | Evidence |
|---|---|---|---|---|
| GGUF Q4_K_M (Hugging Face) gguf-q4_k_m | smollm2-1.7b-instruct-q4_k_m.gguf | 1,055,609,536 (0.98 GiB / 1.06 GB) | llama.cpp, ollama, unspecified | file-metadata |
| GGUF Q5_K_M (Unsloth) gguf-q5_k_m | SmolLM2-1.7B-Instruct-Q5_K_M.gguf | 1,225,478,816 (1.14 GiB / 1.23 GB) | llama.cpp, ollama, unspecified | file-metadata |
| GGUF Q6_K (Unsloth) gguf-q6_k | SmolLM2-1.7B-Instruct-Q6_K.gguf | 1,405,964,960 (1.31 GiB / 1.41 GB) | llama.cpp, ollama, unspecified | file-metadata |
| GGUF Q8_0 (Unsloth) gguf-q8_0 | SmolLM2-1.7B-Instruct-Q8_0.gguf | 1,820,414,624 (1.70 GiB / 1.82 GB) | llama.cpp, ollama, unspecified | file-metadata |
| INT4 theoretical | No published file | 855,688,192 (theoretical) | — | theoretical |
Memory by quantization and context
Concurrent sequences = 1. KV defaults to FP16. The range includes documented runtime extras for the listed engine. These figures use the same core as the interactive calculator.
| Quantization | 2,048 tokens | 8,192 tokens | 32,768 tokens |
|---|---|---|---|
| GGUF Q4_K_M (Hugging Face) | 1.48 GiB–1.89 GiB | 2.61 GiB–3.11 GiB | 7.11 GiB–9.11 GiB |
| GGUF Q5_K_M (Unsloth) | 1.64 GiB–2.05 GiB | 2.77 GiB–3.27 GiB | 7.27 GiB–9.30 GiB |
| GGUF Q6_K (Unsloth) | 1.81 GiB–2.22 GiB | 2.93 GiB–3.44 GiB | 7.43 GiB–9.49 GiB |
| GGUF Q8_0 (Unsloth) | 2.20 GiB–2.61 GiB | 3.32 GiB–3.83 GiB | 7.82 GiB–9.95 GiB |
| INT4 theoretical Theoretical rows are not downloadable artifacts. | 1.30 GiB–2.04 GiB | 2.42 GiB–3.45 GiB | 6.92 GiB–8.98 GiB |
Raw KV contribution
Standard GQA/MHA raw KV is 2 × layers × kvHeads × headDim × tokens × sequences × bytes per element. This page uses KV heads (32), not query heads (32).
- 8,192 tokens, 1 sequence, FP16: 1,610,612,736 bytes (1.5000 GiB)
- 32,768 tokens: 6,442,450,944 bytes (6.00 GiB)
Related pages
Sources
- SmolLM2-1.7B-Instruct model card
- config.json — revision 31b70e2e869a7173562077fd711b654946d38674
- Artifact repo HuggingFaceTB/SmolLM2-1.7B-Instruct-GGUF
- Artifact repo unsloth/SmolLM2-1.7B-Instruct-GGUF
Assumptions that affect this result
- Requested context is within the recorded native window of 8192 tokens.
- Full-resident weight estimates use total parameters, including inactive MoE experts.
- Weight bytes come from the selected weight files only, not from every file in the repository.
- Download file size is not peak GPU memory. Runtime layout, allocator padding, and KV are extra.
- Counted files: smollm2-1.7b-instruct-q4_k_m.gguf.
- Context budget is the number of cached tokens per sequence (prompt plus reserved generation).
- Concurrent sequences share one model replica; weights are not multiplied by sequence count.
- KV dtype is independent of weight quantization unless you change it.
Calculator
Loading catalog snapshot…
Search or pick a catalog checkpoint to see memory demand.
Search or pick a catalog checkpoint to see memory demand.