Skip to content
Run LLM Locally

Qwen3.5 4B on the RTX 5090

With 32 GB of VRAM, the RTX 5090 runs Qwen3.5 4B at Q4_K_M entirely on the GPU with an 8K context (about 4.1 GB needed). At Q8_0 it fits entirely.

ShareFacebookXWhatsAppLinkedInEmail

Qwen3.5 4B

Q4_K_M · 8K context

Excellent

The model needs about 4.1 GB and fits in the 31 GB available, with 26.9 GB to spare.

Your hardware
RTX 5090 32 GB · 64 GB RAM
Estimated memory
~4.1 GB estimate
VRAM (usable memory)
31 GB
GPU only
Yes
CPU/GPU offload
Not required
Generation speed
249–439 tok/s estimate
Largest context in fast memory
256K tokens
Confidence: low

Based on: memory bandwidth (1,792 GB/s) and model size

Memory breakdown

~4.1 GB (3.7 GB–4.6 GB) needed · estimate

Fits entirely in VRAM · 26.9 GB left

  • Model weights2.6 GB
  • KV cache0.3 GB
  • Runtime overhead1 GB
  • Safety margin0.3 GB
Assumptions: context 8K tokens · KV cache f16 · llama.cpp (CUDA) · flash attention on · weights = real file size

EstimateData updated 40 days agoSource: Run LLM Locally engine, Hugging Face config.json + model card

What makes this estimate less reliable

  • no benchmark for this hardware and model yet

Other quantizations on the RTX 5090

Qwen3.5 4B on RTX 5090 + 64 GB RAM · 8K context
Qwen3.5 4B Q8_08K~5.7 GB · 190–306 tok/s estimateExcellent
Qwen3.5 4B Q6_K8K~4.8 GB · 218–367 tok/s estimateExcellent
Qwen3.5 4B Q5_K_M8K~4.5 GB · 232–399 tok/s estimateExcellent
Qwen3.5 4B Q4_K_M8K~4.1 GB · 249–439 tok/s estimateExcellent
Qwen3.5 4B IQ4_XS8K~3.9 GB · 262–469 tok/s estimateExcellent
Qwen3.5 4B Q3_K_M8K~3.7 GB · 271–494 tok/s estimateExcellent

Alternatives

GPUs that run Qwen3.5 4B Q4_K_M entirely in VRAM

Other models that fit the RTX 5090

Where to buy a RTX 5090

Prices and links follow the market you choose.

GeForce RTX 5090

32 GB GDDR7 · 1,792 GB/s · 575 W

ExcellentRuns Qwen3.5 4B Q4_K_M entirely in memory

Price not available

Availability unknown

View on amazon.com

Affiliate link

No current offer imported for this market: the link opens a marketplace search.

Frequently asked questions

Can the RTX 5090 run Qwen3.5 4B?
The model needs about 4.1 GB and fits in the 31 GB available, with 26.9 GB to spare.
Which quantization should I use on the RTX 5090?
Q8_0 is the highest-precision quantization that fits entirely in VRAM at 8K context. ~5.7 GB / 31 GB.
What is the largest context on the RTX 5090?
At Q4_K_M, about 256K tokens fit entirely in VRAM with an f16 KV cache. A q8_0 KV cache roughly doubles that.
How fast is Qwen3.5 4B (RTX 5090)?
We estimate 249–439 tok/s (estimate, confidence low). Based on: memory bandwidth (1,792 GB/s) and model size.