Qwen2.5-Coder-1.5B-Instruct Q4_K_M needs 1.80 GiB
Needs 1.80 GiB: Qwen2.5-Coder-1.5B-Instruct Q4_K_M at 8,192 tokens is an estimate of 1.38 GiB–1.80 GiB. Checkpoint Qwen/Qwen2.5-Coder-1.5B-Instruct. File is 1,117,320,768 bytes. No device is selected, so this is demand only, not a fit.
Specific checkpoints with sourced architecture and weight files. GGUF Q4_K_M (Qwen): 1.38 GiB–1.80 GiB at 8,192 tokens, 1 sequence, FP16 KV, llama.cpp.
Publisher and provenance
- PublisherQwen
- Hugging Face
Qwen/Qwen2.5-Coder-1.5B-Instruct - Revision
2e1fd397ee46e1388853d2af2c993145b0f1098a - Total parameters1,543,714,304
- Layers28
- Query heads12
- KV heads2 (gqa)
- Head dimension128
- Recorded context capability32,768
- 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
- 1.04 GiB (1,117,320,768 bytes)
- Raw KV cache
- 0.22 GiB (234,881,024 bytes)
- Engine extras / workspace
- 0.13 GiB–0.54 GiB
- Total estimated demand
- 1.38 GiB–1.80 GiB
Published artifacts
| Quantization | Selected files | Bytes | Engine | Evidence |
|---|---|---|---|---|
| GGUF Q4_K_M (Qwen) gguf-q4_k_m | qwen2.5-coder-1.5b-instruct-q4_k_m.gguf | 1,117,320,768 (1.04 GiB / 1.12 GB) | llama.cpp, ollama, unspecified | file-metadata |
| GGUF Q5_K_M (Qwen) gguf-q5_k_m | qwen2.5-coder-1.5b-instruct-q5_k_m.gguf | 1,285,494,336 (1.20 GiB / 1.29 GB) | llama.cpp, ollama, unspecified | file-metadata |
| GGUF Q6_K (Qwen) gguf-q6_k | qwen2.5-coder-1.5b-instruct-q6_k.gguf | 1,464,178,752 (1.36 GiB / 1.46 GB) | llama.cpp, ollama, unspecified | file-metadata |
| GGUF Q8_0 (Qwen) gguf-q8_0 | qwen2.5-coder-1.5b-instruct-q8_0.gguf | 1,894,532,160 (1.76 GiB / 1.89 GB) | llama.cpp, ollama, unspecified | file-metadata |
| INT4 theoretical | No published file | 771,857,152 (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 (Qwen) | 1.22 GiB–1.62 GiB | 1.38 GiB–1.80 GiB | 2.04 GiB–2.50 GiB |
| GGUF Q5_K_M (Qwen) | 1.38 GiB–1.78 GiB | 1.54 GiB–1.96 GiB | 2.20 GiB–2.66 GiB |
| GGUF Q6_K (Qwen) | 1.54 GiB–1.95 GiB | 1.71 GiB–2.12 GiB | 2.36 GiB–2.83 GiB |
| GGUF Q8_0 (Qwen) | 1.94 GiB–2.36 GiB | 2.11 GiB–2.53 GiB | 2.76 GiB–3.24 GiB |
| INT4 theoretical Theoretical rows are not downloadable artifacts. | 0.90 GiB–1.57 GiB | 1.06 GiB–1.77 GiB | 1.72 GiB–2.58 GiB |
Raw KV contribution
Standard GQA/MHA raw KV is 2 × layers × kvHeads × headDim × tokens × sequences × bytes per element. This page uses KV heads (2), not query heads (12). Using query heads would overstate this cache by 6×.
- 8,192 tokens, 1 sequence, FP16: 234,881,024 bytes (0.2188 GiB)
- 32,768 tokens: 939,524,096 bytes (0.8750 GiB)
Related pages
Sources
- Qwen2.5-Coder-1.5B-Instruct model card
- config.json — revision 2e1fd397ee46e1388853d2af2c993145b0f1098a
- Artifact repo Qwen/Qwen2.5-Coder-1.5B-Instruct-GGUF
Assumptions that affect this result
- Requested context is within the recorded native window of 32768 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: qwen2.5-coder-1.5b-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.