Skip to content
Run LLM Locally

Mistral Nemo 12B Instruct on the RTX 5090

With 32 GB of VRAM, the RTX 5090 runs Mistral Nemo 12B Instruct at Q4_K_M entirely on the GPU with an 8K context (about 9.4 GB needed). At Q8_0 it fits entirely.

ShareFacebookXWhatsAppLinkedInEmail

Mistral Nemo 12B Instruct

Q4_K_M · 8K context

Excellent

The model needs about 9.4 GB and fits in the 31 GB available, with 21.6 GB to spare.

Your hardware
RTX 5090 32 GB · 64 GB RAM
Estimated memory
~9.4 GB estimate
VRAM (usable memory)
31 GB
GPU only
Yes
CPU/GPU offload
Not required
Generation speed
121–185 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

~9.4 GB (9 GB–10 GB) needed · estimate

Fits entirely in VRAM · 21.6 GB left

  • Model weights7 GB
  • KV cache1.3 GB
  • Runtime overhead0.8 GB
  • Safety margin0.4 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

Mistral Nemo 12B Instruct on RTX 5090 + 64 GB RAM · 8K context
Mistral Nemo 12B Instruct Q8_08K~14.9 GB · 81–117 tok/s estimateExcellent
Mistral Nemo 12B Instruct Q6_K8K~12 GB · 99–146 tok/s estimateExcellent
Mistral Nemo 12B Instruct Q5_K_M8K~10.7 GB · 109–164 tok/s estimateExcellent
Mistral Nemo 12B Instruct Q4_K_M8K~9.4 GB · 121–185 tok/s estimateExcellent
Mistral Nemo 12B Instruct IQ4_XS8K~8.7 GB · 129–200 tok/s estimateExcellent
Mistral Nemo 12B Instruct Q3_K_M8K~8.1 GB · 137–217 tok/s estimateExcellent

Alternatives

GPUs that run Mistral Nemo 12B 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 Mistral Nemo 12B 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 Mistral Nemo 12B Instruct?
The model needs about 9.4 GB and fits in the 31 GB available, with 21.6 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. ~14.9 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 Mistral Nemo 12B Instruct (RTX 5090)?
We estimate 121–185 tok/s (estimate, confidence low). Based on: memory bandwidth (1,792 GB/s) and model size.