Yes — with 7.3 GB to spare
Llama 3.2 3B Instruct at Q4_K_M fits your GeForce RTX 5070 entirely on the GPU at 8K context, at an estimated 225 tokens per second. Past 74K the KV cache pushes it over — quantise the cache to q8_0, or step down a quantisation, to go longer.
Fully on GPU
8K context
Q4_K_M · 1.8 GB
Llama 3.2 Community
Released Sep 2024
The 2024 "it just runs" model for 8 GB laptops. Qwen3.5 4B does the same job better now.
The VRAM budget
weights 1.8 GB
Weights 1.8 GB
KV cache @ 8K 0.88 GB
Runtime overhead 0.6 GB
Free 7.3 GB of 10.6 GB
Quantisation ladder
| Quant | Weights | Total @ 8K | Max context | Tok/s | Quality | Fit |
|---|---|---|---|---|---|---|
| F16 | 6.0 GB | 7.5 GB | 36K | 68 | Reference | Long context |
| Q8_0 | 3.2 GB | 4.7 GB | 62K | 128 | −0.1% ppl | Long context |
| Q6_K | 2.5 GB | 3.9 GB | 69K | 166 | −0.4% ppl | Long context |
| Q5_K_M | 2.1 GB | 3.6 GB | 72K | 192 | −0.8% ppl | Long context |
| Q4_K_M | 1.8 GB | 3.3 GB | 74K | 225 | −1.9% ppl | Recommended |
| Q3_K_M | 1.5 GB | 2.9 GB | 78K | 278 | −5.4% ppl | Long context |
| Q2_K | 1.3 GB | 2.7 GB | 79K | 325 | −15% ppl | Long context |
Quality is the published perplexity delta against f16 weights. Max context assumes an f16 KV cache; q8_0 roughly doubles it.
How to run it
$ ollama pull llama3.2:3b $ OLLAMA_CONTEXT_LENGTH=8192 \ ollama run llama3.2:3b
The default. One binary, a model registry, an OpenAI-compatible port. More on Ollama.
01Download is 1.8 GB. Keep it on an SSD — a first load off a spinning disk takes minutes.
02Close anything else holding VRAM. A browser with hardware acceleration can sit on 1–2 GB.
03There is room to go to 74K context on this card.