Skip to content
Run LLM Locally

Llama 3.2 1B Instruct on the RTX 5090

With 32 GB of VRAM, the RTX 5090 runs Llama 3.2 1B Instruct at Q4_K_M entirely on the GPU with an 8K context (about 2 GB needed). At Q8_0 it fits entirely.

ShareFacebookXWhatsAppLinkedInEmail

Llama 3.2 1B Instruct

Q4_K_M · 8K context

Excellent

The model needs about 2 GB and fits in the 31 GB available, with 29 GB to spare.

Your hardware
RTX 5090 32 GB · 64 GB RAM
Estimated memory
~2 GB estimate
VRAM (usable memory)
31 GB
GPU only
Yes
CPU/GPU offload
Not required
Generation speed
610–1,174 tok/s estimate
Largest context in fast memory
128K tokens
Confidence: low

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

Memory breakdown

~2 GB (1.7 GB–2.3 GB) needed · estimate

Fits entirely in VRAM · 29 GB left

  • Model weights0.8 GB
  • KV cache0.3 GB
  • Runtime overhead0.7 GB
  • Safety margin0.3 GB
Assumptions: context 8K tokens · KV cache f16 · llama.cpp (CUDA) · flash attention on · weights = real file size

EstimateData updated 24 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

Llama 3.2 1B Instruct on RTX 5090 + 64 GB RAM · 8K context
Llama 3.2 1B Instruct Q8_08K~2.5 GB · 497–876 tok/s estimateExcellent
Llama 3.2 1B Instruct Q6_K8K~2.2 GB · 557–1,028 tok/s estimateExcellent
Llama 3.2 1B Instruct Q5_K_M8K~2.1 GB · 583–1,098 tok/s estimateExcellent
Llama 3.2 1B Instruct Q4_K_M8K~2 GB · 610–1,174 tok/s estimateExcellent
Llama 3.2 1B Instruct IQ4_XS8K~2 GB · 628–1,226 tok/s estimateExcellent
Llama 3.2 1B Instruct Q3_K_M8K~1.9 GB · 643–1,272 tok/s estimateExcellent

Alternatives

GPUs that run Llama 3.2 1B Instruct 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 Llama 3.2 1B Instruct 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 Llama 3.2 1B Instruct?
The model needs about 2 GB and fits in the 31 GB available, with 29 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. ~2.5 GB / 31 GB.
What is the largest context on the RTX 5090?
At Q4_K_M, about 128K tokens fit entirely in VRAM with an f16 KV cache. A q8_0 KV cache roughly doubles that.
How fast is Llama 3.2 1B Instruct (RTX 5090)?
We estimate 610–1,174 tok/s (estimate, confidence low). Based on: memory bandwidth (1,792 GB/s) and model size.