Qwen2.5 7B Instruct: memory fit on every Jetson module.
Method v1.0 · dataset 2026-09-07 · verified 2026-09-07 · last updated September 2026
7.61 B parameters · 28 layers · 32,768-token max context. Hugging Face: Qwen/Qwen2.5-7B-Instruct.
Model facts
| Fact | Value | Note |
|---|---|---|
| Parameters | 7.61 B | |
| Layers | 28 | |
| Hidden size | 3584 | |
| Attention heads | 28 | |
| KV heads | 4 | |
| Head dimension | 128 | |
| Max context | 32,768 tokens | |
| dtype | bfloat16 | |
| Vocabulary | 152,064 |
Qwen (Alibaba) · hf:Qwen/Qwen2.5-7B-Instruct · config.json · verified 2026-09-07
“"hidden_size": 3584, "num_attention_heads": 28, "num_key_value_heads": 4”
Parameters source
- Qwen (Alibaba) · hf:Qwen/Qwen2.5-7B-Instruct · verified 2026-09-07 · Qwen/Qwen2.5-7B-Instruct README.md · class A
“Number of Parameters: 7.61B”
Published quantised artefacts
A file the publisher or a community mirror actually ships, used as the class-A weight figure when the requested quant matches. Any quant without a row here falls back to parameters × bytes-per-parameter (class D for block quants).
| Quant | Format | File | Size | Source |
|---|---|---|---|---|
| GGUF Q4_K_M / INT4 (AWQ, MLC q4f16) | GGUF Q4_K_M | Qwen2.5-7B-Instruct-Q4_K_M.gguf | 4.683 GB | bartowski (community GGUF) · hf:bartowski/Qwen2.5-7B-Instruct-GGUF · verified 2026-09-07 |
Jetson tokens/s measurements
Published or archived throughput numbers, not modelled. Class C (external measured benchmark).
| Module | Runtime | Quant | Measured | Source |
|---|---|---|---|---|
| Jetson Orin Nano Super | unspecified (archived benchmark) | — | 21.75 tok/s | NVIDIA Jetson AI Lab (archive) · jetson-ai-lab:benchmarks.html · verified 2026-09-07 |
Every module × every quantisation
Verdict and total memory at 4096-token context, 1 sequence, llama.cpp, headless. Q4 cells link to the static breakdown page; every other cell links to the live engine at that quantisation.
Notes
head_dim derived as hidden_size/num_attention_heads=3584/28=128.
Method and data: Model Memory Fit methodology. Full registry: model-memory.json.
Change the context, concurrency or runtime.
The live engine covers any context length, sequence count, KV precision and runtime, with a permanent link.