Yes — with 18.9 GB to spare
Qwen3.5 4B at Q4_K_M fits your Tesla P40 entirely on the GPU at 8K context, at an estimated 80 tokens per second. There is room for its full 256K window.
Fully on GPU
8K context
Q4_K_M · 2.6 GB
Apache 2.0
Released 28 Feb 2026
Vision
The 8 GB coding agent. Q4 lands near 3.4 GB, leaving room for a real context window.
The VRAM budget
weights 2.6 GB
Weights 2.6 GB
KV cache @ 8K 0.25 GB
Runtime overhead 0.6 GB
Free 18.9 GB of 22.4 GB
Quantisation ladder
| Quant | Weights | Total @ 8K | Max context | Tok/s | Quality | Fit |
|---|---|---|---|---|---|---|
| F16 | 8.7 GB | 9.5 GB | 256K | 24 | Reference | Long context |
| Q8_0 | 4.6 GB | 5.5 GB | 256K | 45 | −0.1% ppl | Long context |
| Q6_K | 3.6 GB | 4.4 GB | 256K | 59 | −0.4% ppl | Long context |
| Q5_K_M | 3.1 GB | 3.9 GB | 256K | 68 | −0.8% ppl | Long context |
| Q4_K_M | 2.6 GB | 3.5 GB | 256K | 80 | −1.9% ppl | Recommended |
| Q3_K_M | 2.1 GB | 3.0 GB | 256K | 99 | −5.4% ppl | Long context |
| Q2_K | 1.8 GB | 2.7 GB | 256K | 115 | −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. Only 8 of its 32 blocks keep a per-token KV cache; the rest are linear-attention, Mamba or convolution blocks with a fixed-size state.
How to run it
$ ollama pull qwen3.5:4b $ OLLAMA_CONTEXT_LENGTH=8192 \ ollama run qwen3.5:4b
The default. One binary, a model registry, an OpenAI-compatible port. More on Ollama.
01Download is 2.6 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 the model's full 256K context on this card.