Magistral Small 2506 — Hardware Requirements & GPU Compatibility
Chat
Magistral Small 2506 is a 23.6B-parameter open language model from Mistral AI. It supports a context window of up to 40,960 tokens. At BF16 it needs about 47.86 GB of VRAM — see which GPUs and Macs can run it below.
48.6K downloads
608 likesSpecifications
- Publisher
- Mistral AI
- Parameters
- 23.6B
- Architecture
- MistralForCausalLM
- Context Length
- 40,960 tokens
- Vocabulary Size
- 131,072
- License
- Apache 2.0
Get Started
HuggingFace
How Much VRAM Does Magistral Small 2506 Need?
Select a quantization to see compatible GPUs below.
| Quantization | Bits | VRAM | + Context | File Size | Quality |
|---|---|---|---|---|---|
| BF16 | 16.00 | 47.9 GB | 55.8 GB | 47.14 GB | Brain floating point 16 — preferred for training |
Which GPUs Can Run Magistral Small 2506?
BF16 · 47.9 GBMagistral Small 2506 (BF16) requires 47.9 GB of VRAM to load the model weights. For comfortable inference with headroom for KV cache and system overhead, 63+ GB is recommended. Using the full 41K context window can add up to 8.0 GB, bringing total usage to 55.8 GB. No single GPU has enough memory — multi-GPU or cluster setups are needed.
Which Devices Can Run Magistral Small 2506?
BF16 · 47.9 GB11 devices with unified memory can run Magistral Small 2506, including NVIDIA DGX H100, NVIDIA DGX A100 640GB, Mac Studio M4 Max (64 GB).
Runs great
— Plenty of headroom
Benchmarks
View all 3 →Related Models
Frequently Asked Questions
- How much VRAM does Magistral Small 2506 need?
Magistral Small 2506 requires 47.9 GB of VRAM at BF16. Full 41K context adds up to 8.0 GB (55.8 GB total).
VRAM = Weights + KV Cache + Overhead
Weights = 23.6B × 16 bits ÷ 8 = 47.1 GB
KV Cache + Overhead ≈ 0.8 GB (at 2K context + ~0.3 GB framework)
KV Cache + Overhead ≈ 8.7 GB (at full 41K context)
VRAM usage by quantization
BF1647.9 GBBF16 + full context55.8 GB- Can NVIDIA GeForce RTX 5090 run Magistral Small 2506?
No — Magistral Small 2506 requires at least 47.9 GB at BF16, which exceeds the NVIDIA GeForce RTX 5090's 32 GB of VRAM.
- Can I run Magistral Small 2506 on a Mac?
Magistral Small 2506 requires at least 47.9 GB at BF16, which exceeds the unified memory of most consumer Macs. You would need a Mac Studio or Mac Pro with a high-memory configuration.
- Can I run Magistral Small 2506 locally?
Yes — Magistral Small 2506 can run locally on consumer hardware. At BF16 quantization it needs 47.9 GB of VRAM. Popular tools include Ollama, LM Studio, and llama.cpp.
- How fast is Magistral Small 2506?
At BF16, Magistral Small 2506 can reach ~61 tok/s on AMD Instinct MI300X. Speed depends mainly on GPU memory bandwidth. Real-world results typically within ±20%.
tok/s = (bandwidth GB/s ÷ model GB) × efficiency
Example: AMD Instinct MI300X → 5300 ÷ 47.9 × 0.55 = ~61 tok/s
Estimated speed at BF16 (47.9 GB)
~61 tok/s~46 tok/s~38 tok/sReal-world results typically within ±20%. Speed depends on batch size, quantization kernel, and software stack.
- What's the download size of Magistral Small 2506?
At BF16, the download is about 47.14 GB.
- Which GPUs can run Magistral Small 2506?
No single consumer GPU has enough VRAM to run Magistral Small 2506 at BF16 (47.9 GB). Multi-GPU or professional hardware is required.
- Which devices can run Magistral Small 2506?
11 devices with unified memory can run Magistral Small 2506 at BF16 (47.9 GB), including Mac Mini M4 Pro (48 GB), Mac Pro M2 Ultra (192 GB), Mac Studio M2 Ultra (192 GB), Mac Studio M4 Max (128 GB). Apple Silicon Macs use unified memory shared between CPU and GPU, making them well-suited for local LLM inference.