LLM GPU requirements
Mistral Small 24B GPU requirements
How much GPU memory Mistral Small 24B needs in BF16, FP8 and 4-bit, how its KV cache grows with context, what fine-tuning takes, and which servers run it, with monthly prices.
- Parameters
- 23.6B
- Architecture
- Dense 40 layers
- KV heads
- 8 × 128 dim
- Context
- 128k tokens
Cheapest servers that run it
FP8 weights · 8k context · 8 concurrent requests · about 38 GB
-
NVIDIA L40S In stock
48 GB GDDR6 · Single GPU
$890/mo
-
NVIDIA RTX PRO 6000 Blackwell In stock
96 GB GDDR7 · Single GPU
$1,390/mo
-
NVIDIA H100 SXM In stock
80 GB HBM3 · Single GPU
$2,190/mo
Delivery date confirmed before you pay. Missed it? Full refund.
Inference memory
How much VRAM Mistral Small 24B needs
Weights take parameters × bytes per parameter. Serving adds the KV cache and about 10% runtime overhead; the totals below assume an 8k context with 8 concurrent requests and a BF16 KV cache.
| Precision | Weights | Total to serve | Cheapest server that fits |
|---|---|---|---|
| BF16 2 bytes per parameter | 47 GB | 64 GB | NVIDIA RTX PRO 6000 Blackwell $1,390/mo |
| FP8 1 byte per parameter | 24 GB | 38 GB | NVIDIA L40S $890/mo |
| 4-bit 0.5 bytes per parameter | 12 GB | 25 GB | NVIDIA GeForce RTX 5090 $449/mo |
Estimates with about 8% of GPU memory kept free. Quantized weights also keep some layers in higher precision, and frameworks differ in overhead; check the free memory your engine reports at startup. A model split across GPUs with tensor parallelism runs best on NVLink servers (HGX); the 8× RTX PRO 6000 server links its GPUs over PCIe.
KV cache
Mistral Small 24B KV cache by context length
Each token in flight stores keys and values for every layer: 2 × 40 layers × 8 KV heads × 128 dimensions × 2 bytes = 160 KB per token in BF16. An FP8 KV cache halves these figures.
| Context per request | 1 request | 8 requests | 32 requests |
|---|---|---|---|
| 4k tokens | 0.7 GB | 5.4 GB | 21 GB |
| 32k tokens | 5.4 GB | 43 GB | 172 GB |
| 128k tokens | 21 GB | 172 GB | 687 GB |
Fine-tuning
Memory to fine-tune Mistral Small 24B
Rough totals before tuning batch size and sequence length: QLoRA keeps 4-bit base weights, LoRA keeps BF16 weights frozen, and a full fine-tune with AdamW needs about 16 bytes per parameter plus activations, sharded across GPUs.
| Method | Approximate memory | Cheapest server that fits |
|---|---|---|
| QLoRA (4-bit base) | 18 GB | NVIDIA GeForce RTX 5090 $449/mo |
| LoRA (BF16 base) | 59 GB | NVIDIA RTX PRO 6000 Blackwell $1,390/mo |
| Full fine-tune (AdamW, mixed precision) | 453 GB | 8× NVIDIA RTX PRO 6000 Blackwell $10,490/mo |
Full fine-tuning assumes sharded training (FSDP or ZeRO-3) with activation checkpointing. See the fine-tuning guide for methods and settings.
About Mistral Small 24B
Mistral Small 3.x is a 24B dense model family (about 23.6B parameters) with 40 layers and 8 KV heads, released under Apache 2.0; versions 3.1 and later accept a 128k-token context. It sits between the 14B and 32B classes: in BF16 its weights take about 47 GB, and FP8 halves that.
Every figure on this page comes from the model’s published architecture and the same formulas as our VRAM calculator, where you can change the context length, concurrency and KV cache precision.
Model facts
- Developer
- Mistral AI
- License
- Apache 2.0
- Type
- Dense
- Parameters
- 23.6B
- Layers
- 40
- Attention
- 32 query heads, 8 KV headsgrouped-query attention, head dimension 128
- Context window
- 128k tokens (3.1 and later)
FAQ
Mistral Small 24B GPU FAQ
The short version. Anything else, our engineers reply 24/7.
Ask an engineerHow much VRAM does Mistral Small 24B need?
About 47 GB for the weights in BF16, 24 GB in FP8 and 12 GB in 4-bit, plus the KV cache: about 0.2 GB per 1,000 tokens in flight with a BF16 cache. Serving it in FP8 with an 8k context and 8 concurrent requests takes about 38 GB in total.
Can Mistral Small 24B run on a single GPU?
Yes. In FP8, a single NVIDIA L40S (48 GB) serves it with an 8k context and 8 concurrent requests. Longer contexts or more users need more memory for the KV cache, so check your case in the VRAM calculator.
What is the cheapest FullGPU server for Mistral Small 24B?
For FP8 serving with an 8k context and 8 concurrent requests, the NVIDIA L40S at $890 per month is the lowest-priced configuration that fits. For other contexts and concurrency, size it in the VRAM calculator or ask an engineer.
How much memory do I need to fine-tune Mistral Small 24B?
Roughly 18 GB with QLoRA, 59 GB with LoRA on BF16 weights, and 453 GB for a full fine-tune with AdamW in mixed precision, before tuning batch size and sequence length. Full fine-tunes of large models are sharded across GPUs with FSDP or ZeRO-3.
Talk to an engineer
Run Mistral Small 24B on your own GPUs.
Tell us what you need. A GPU engineer replies with availability, a confirmed delivery date and a payment link.
- Delivery date confirmed before you pay
- Cancel any month
- Card or crypto, same price