Mistral Nemo 12B VRAM: what it needs at Q4_K_M, Q8_0 and FP16
Runs in your browser — nothing you paste leaves this page. How we prove that
LLM VRAM calculator playground
KV cache is counted at FP16. Overhead is a fixed 1.5 GiB estimate for the runtime and compute buffers. A custom size borrows the architecture of the nearest preset.
Results update as you type — press Enter to run now.
GPU tiers
- 8 GiBtoo smallRTX 4060, RTX 3070, RX 7600
- 12 GiBfitsRTX 4070, RTX 3060 12GB, RTX 5070
- 16 GiBfitsRTX 4060 Ti 16GB, RTX 4080, RTX 5080
- 24 GiBfitsRTX 4090, RTX 3090, RX 7900 XTX
- 32 GiBfitsRTX 5090, V100 32GB
- 48 GiBfitsRTX A6000, RTX 6000 Ada, L40S
- 80 GiBfitsA100 80GB, H100 80GB
At Q4_K_M with an 8,192-token context, Mistral Nemo 12B needs about 9.58 GiB of VRAM: 6.83 GiB of weights, 1.25 GiB of KV cache and 1.50 GiB of runtime overhead. The smallest common GPU tier that holds it is 12 GiB (RTX 4070, RTX 3060 12GB, RTX 5070).
How much VRAM Mistral Nemo 12B needs
Mistral Nemo 12B, built by Mistral AI with NVIDIA and released in July 2024 under Apache 2.0, offers a 128k context and the Tekken tokenizer at a size that suits 12 and 16 GiB cards.
With 8,192 tokens of context the estimate is 25.1 GiB at FP16, 14.6 GiB at Q8_0 and 9.58 GiB at Q4_K_M. FP16 needs a 32 GiB card, Q8_0 needs a 16 GiB card, and Q4_K_M needs a 12 GiB card. The weights are 12 billion parameters times the bits per weight of each format (16, 8.5 and 4.89), divided by eight; everything is in GiB, the unit GPU memory is sold in.
Context length and the KV cache
The KV cache stores a key and a value vector for every token in every layer. Mistral Nemo 12B has 40 layers, 8 KV heads and a head dimension of 128, read from its Hugging Face config.json, so at FP16 each token costs 160 KiB. That is 1.25 GiB at 8,192 tokens, and 20.0 GiB at the maximum of 131,072, where the Q4_K_M total becomes 28.3 GiB and needs a 32 GiB card. The cache does not shrink with weight quantization; only a runtime option that quantizes the cache itself does.
Which quantization to pick
Q4_K_M at 9.58 GiB fits a 12 GiB card; Q5_K_M at 10.7 GiB still fits 12 GiB, and Q8_0 at 14.6 GiB needs a 16 GiB card.
Pitfalls
Its head dimension is 128, set explicitly in the config rather than derived from the hidden size, which is why calculators that divide hidden size by heads overestimate its cache. The full 131,072 tokens add 20.0 GiB of cache. Mistral recommends a lower sampling temperature than most models, around 0.3.
Compare Mistral 7B v0.3 for less memory, or Qwen2.5 14B and Gemma 2 9B at a similar tier.
FAQ
Questions, answered.
Tap a question to expand the answer.
Can Mistral Nemo 12B run on a 12 GB GPU?
Yes at Q4_K_M: 9.58 GiB at 8,192 tokens and 10.8 GiB at 16,384. At 32,768 tokens it needs 13.3 GiB, which calls for a 16 GiB card.
How much VRAM does Mistral Nemo need for its full 128k context?
28.3 GiB at Q4_K_M, with 20.0 GiB of KV cache, so a 32 GiB card.
More free, private DevOps tools.
The LLM VRAM Calculator is one tool in OpsCanopy — a growing canopy of browser-based validators, converters and testers that never touch a server.
Related tools
New to AI & local LLMs? Read the AI & local LLMs guide →
42 free tools, every one offline-capable — opscanopy.com works with no signup and nothing uploaded.
Related: the full LLM VRAM Calculator, the LLM Token Counter, or every AI tool.
Estimates only; the KV cache is sized at FP16 and the overhead is a fixed allowance, so always confirm against your own runtime’s report before buying hardware. OpsCanopy is free and open.