VILA 1.5 8B (Llama-3): memory fit on every Jetson module.
Method v1.0 · dataset 2026-09-07 · verified 2026-09-07 · last updated September 2026
8.49 B parameters (text 8.06 B + vision tower) · 32 layers · 8,192-token max context. Hugging Face: Efficient-Large-Model/Llama-3-VILA1.5-8B.
Model facts
| Fact | Value | Note |
|---|---|---|
| Parameters | 8.494 B | text 8.064 B + vision 0.43 B (estimated) |
| Layers | 32 | |
| Hidden size | 4096 | |
| Attention heads | 32 | |
| KV heads | 8 | |
| Head dimension | 128 | |
| Max context | 8,192 tokens | |
| dtype | bfloat16 | |
| Vocabulary | 128,257 |
Efficient-Large-Model (VILA) · hf:Efficient-Large-Model/Llama-3-VILA1.5-8B · config.json · verified 2026-09-07
“"hidden_size": 4096, "num_attention_heads": 32, "num_key_value_heads": 8, "num_hidden_layers": 32”
Parameters source
- Hugging Face · hf:Efficient-Large-Model/Llama-3-VILA1.5-8B · verified 2026-09-07 · repo blob listing (llm/, vision_tower/, mm_projector/ safetensors sizes) · class A
“llm/model-0000{1..4}-of-00004.safetensors sizes 4976706864 + 4999802720 + 4915916176 + 1168147000, vision_tower/model.safetensors 856506120, mm_projector/model.safetensors 71338512”
Vision tower
| Fact | Value |
|---|---|
| Layers | 27 |
| Hidden size | 1152 |
| Patch size | 14 |
| Image size | 384 |
| Model type | siglip_vision_model |
Vision config source
- Efficient-Large-Model (VILA) · hf:Efficient-Large-Model/Llama-3-VILA1.5-8B · vision_config · verified 2026-09-07 · vision_tower/config.json · class A
“"model_type": "siglip_vision_model", "hidden_size": 1152, "num_hidden_layers": 27”
Image tokens: 196 tokens per image (SigLIP 384 / 14 ≈ 27² patches downsampled ~4× by the VILA projector; 196 assumed; class E).
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).
No published quantised artefact for this model; weights are estimated as parameters × bytes-per-parameter for every quantisation.
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) | — | 0.83 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
No GGUF quantization found for VILA -- see gaps. LLM backbone is Llama-3-8B architecture (matches llama-3.1-8b-instruct layer/head config except max_position_embeddings=8192 here vs 131072 for the released Llama 3.1 chat model, and rope_theta 500000).
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.