llmrun
Mistral AIMistralMistral3ForConditionalGeneration

Mistral Small 3.2 24B Instruct 2506 — Hardware Requirements & GPU Compatibility

Chat

Mistral-Small-3.2-24B-Instruct-2506 is a dense 24-billion-parameter vision-language model from Mistral AI, a minor refinement of Mistral-Small-3.1-24B-Instruct-2503. The update focuses on following precise instructions more reliably, cutting down on repetitive or runaway generations, and making function calling more robust, while still handling both text and image inputs for tasks like chart and document understanding. It supports a 131,072-token context window and is released under the Apache 2.0 license. At this size, 4-bit quantization needs roughly 14GB of memory, so it runs on a single consumer GPU in the 16-24GB range, such as an RTX 4090.

24.0B params · dense131K contextApache 2.0260.1K downloads637 likes105.7K quant downloads
A

Runs well on a 24 GB GPU

Runs from12 GB IQ2_M · 16KBest on24 GB Q4_K_M · ~43 tok/s · 32KMacs from16 GB IQ2_M · ~16 tok/s

Score 85.7#73 of 100 open#20 of 35 that fit 24 GBApache 2.0

Reference GPU · RTX 4090 24 GB

Estimates, ±20% · how we estimate

llmrun Composite Score →

85.7

Benchmarks · tier chart ↓
Among 100 open modelsHow the Composite Score is built

Run Mistral Small 3.2 24B Instruct 2506

One recommended way. Every other method, OS and quant is in the step-by-step guide.

Recommended

Ollama
one command · needs Ollama installed
$ollama run mistral-small3.2:24b

Q4_K_M fits fully on the RTX 4090 · 5.9 GB free · 16K context

Download: Mistral-Small-3.2-24B-Instruct-2506-Q4_K_M.gguf14.33 GBunsloth

install, OS, other quants

Other ways

each opens step-by-step

Run in cloud

Fits on RTX 3090 (24 GB) (5 GB headroom) · Q4_K_M

Generation speed
~40 tok/s
generation speed
Cost per 1M output tokens
$1.49
per 1M output tokens
Deployon Vast.aiBest optionfrom $0.22/hr (opens in new tab)
Compare GPUs →
or

How Much VRAM Does Mistral Small 3.2 24B Instruct 2506 Need?

Select a quantization level; the listed hardware adheres to it.

QuantizationVRAM+32KOn RTX 4090 (reference)
10.9 GB17.2 GB2-bit · K-quant — 2-bit quantization with K-quant improvementsFits: 64K context · ~60 tok/s
11.2 GB17.5 GB3-bit small — 3-bit small quantizationFits: 32K context · ~58 tok/s
12.4 GB18.7 GB3-bit medium — 3-bit medium quantizationFits: 32K context · ~53 tok/s
12.7 GB19 GB4-bit · legacy — 4-bit legacy quantizationFits: 32K context · ~52 tok/s
15.1 GB21.4 GB4-bit medium — 4-bit medium quantization — most popular sweet spotFits: 32K context · ~43 tok/s
17.8 GB24.1 GB5-bit medium — 5-bit medium quantization — good quality/size tradeoffFits: 16K context · ~37 tok/s
20.5 GB26.8 GB6-bit, very good — 6-bit quantization, very good qualityFits: 16K context · ~32 tok/s
24.7 GB31 GB8-bit, near-lossless — 8-bit quantization, near-losslessDoes not fit: needs 27.7 GB

Selected Q4_K_M

4-bit medium quantization

18.1GBat 16K context

Weights
14.4 GB
KV cache
3.4 GB
Overhead
0.3 GB

Fits RTX 4090 (reference) · 5.9 GB free · ~43 tok/s

≥50 tok/s needs ≥1.16 TB/s · RTX 5090 ~77

Which Hardware Can Run Mistral Small 3.2 24B Instruct 2506?

Q4_K_M · 15.1 GB

Mistral Small 3.2 24B Instruct 2506 (Q4_K_M) requires 15.1 GB of VRAM to load the model weights. For comfortable inference with headroom for KV cache and system overhead, 20+ GB is recommended. Using the full 131K context window can add up to 26.4 GB, bringing total usage to 41.5 GB. 26 GPUs can run it, including NVIDIA GeForce RTX 5090, NVIDIA GeForce RTX 3090 Ti, NVIDIA GeForce RTX 5080. 122 devices with unified memory can run Mistral Small 3.2 24B Instruct 2506, including Mac Studio (M3 Ultra, 256GB), Mac Studio (M3 Ultra, 512GB), Mac Mini M5 Pro (24 GB).

By VRAM tier · typical GPUs and Macs at Q4_K_M
8 GB · RTX 4060Doesn't fit — smallest quant needs 10.3 GB8 GB tier →
12 GB · RTX 3060 12GBIQ2_M · 11.8 GB · ~27 tok/s · 16K12 GB tier →
16 GB · RTX 5060 Ti 16GBQ4_K_M · 18.1 GB · ~19 tok/s · 4Ktight16 GB tier →
24 GB · RTX 4090referenceQ4_K_M · 18.1 GB · ~43 tok/s · 32K24 GB tier →
Mac 32 GB · Mac Mini M4Q4_K_M · 18.1 GB · ~6 tok/s · 64KMacs →
Mac 64 GB · Mac Studio M4 MaxQ4_K_M · 18.1 GB · ~25 tok/s · 128KMacs →

How good is Mistral Small 3.2 24B Instruct 2506?

The llmrun Score merges 5 public benchmark boards into a single figure, where 100 represents the average for open models.

Score vs VRAM in the 24 GB tier

  • Mistral Small 3.2 24B Instruct 2506
  • Other models
  • Best Score for its VRAM

#20 of 35 that fit 24 GB

Each dot is a model at the best quant that fits a 24 GB GPU. The line joins the models with the best Score for their VRAM; a dot below it is beaten by a model that needs no more memory.

Full 24 GB ranking →

Benchmarks

Here are all 6 boards that have results for this model, sorted by their weight in the llmrun Score; badges show who ran the test.

  • GPQA Diamond

    Epoch AI · Jun 2025Independent

    49.0%

    #55/98 open

    Independent
  • LMArena Text

    LMArena · Oct 2026Arena

    1356.1

    #71/187 open

    Arena
  • AIME 2024/2025

    Epoch AI · Jun 2025Independent

    30.3%

    #49/86 open

    Independent
  • SciCode

    Epoch AI · Jun 2025Lab-reported

    26.4%

    #57/66 open

    Lab-reported
  • CritPt

    Epoch AI · Jun 2025Lab-reported

    0.0%

    #61/66 open

    Lab-reported
  • DTBench

    Epoch AI · Jun 2025Lab-reported

    59.9%

    #49/68 open

    Lab-reported
All 6 Mistral Small 3.2 24B Instruct 2506 benchmarks

In the 24 GB class

35 models fit 24 GB

Same Score, only models that fit 24 GB at their best quant: the ranking a 24 GB owner actually chooses from.

  1. #1

    Qwen3.8 27BAlibaba

    122.1

    ~38 tok/s · 17.9 GB

  2. #2

    Qwen3.6 27BAlibaba

    111.2

    ~38 tok/s · 17.9 GB

  3. #3

    Gemma 4 31B ITGoogle

    106.8

    ~32 tok/s · 21.7 GB

  4. #4

    Qwen3.6 35B A3BAlibaba

    106.8

    ~151 tok/s · 22.0 GB

  5. #5

    Gemma 4 26B A4B ITGoogle

    105.8

    ~171 tok/s · 16.5 GB

  6. #6

    Qwen3.5 35B A3BAlibaba

    105.8

    ~151 tok/s · 22.0 GB

Full 24 GB ranking →RTX 4090 page →

Alternatives to Mistral Small 3.2 24B Instruct 2506

Mistral family · 34 models →

Official models with an llmrun Score, each answering one question; numbers are at the best quant that fits a 24 GB GPU.

Similar models

10
ModelScoreVRAMtok/sWhy it is similar
Gemma 4 31B ITGoogle106.821.7 GB~32similar score
Gemma 4 26B A4B ITGoogle105.816.5 GB~171similar score
Qwen3.5 35B A3BAlibaba105.822 GB~151similar score
Qwen3.5 27BAlibaba103.817.9 GB~38similar score
Seed OSS 36B InstructByteDance-Seed102.522.8 GB~31similar score
Qwen3 30B A3B Thinking 2507Alibaba100.719.4 GB~173similar score
Qwen3.5 9BAlibaba99.46.6 GB~103similar score
Qwen3 14BAlibaba98.011.9 GB~69similar score
DeepSeek R1 Distill Qwen 32BDeepSeek97.223 GB~34similar score
QwQ 32BAlibaba96.223 GB~34similar score
Fine-tunes & variants(4)

Specifications

Publisher
Mistral AI
Family
Mistral
Parameters
24.0B
Architecture
Mistral3ForConditionalGeneration
Context Length
131,072 tokens
Vocabulary Size
131,072
Release Date
2025-06-19
License
Apache 2.0

Resources

Frequently Asked Questions

How much VRAM does Mistral Small 3.2 24B Instruct 2506 need?

Mistral Small 3.2 24B Instruct 2506 requires 15.1 GB of VRAM at Q4_K_M, or 48.7 GB at BF16. Full 131K context adds up to 26.4 GB (41.5 GB total).

VRAM = Weights + KV Cache + Overhead

Weights = 24.0B × 4.8 bits ÷ 8 = 14.4 GB

KV Cache + Overhead ≈ 0.7 GB (at 2K context + ~0.3 GB framework)

Fit ratings and hardware model lists check this model with room for a 16K-token context, which needs a little more memory.

KV Cache + Overhead ≈ 27.2 GB (at full 131K context)

VRAM usage by quantization

15.1 GB
41.5 GB

Learn more about VRAM estimation →

Can NVIDIA GeForce RTX 4090 run Mistral Small 3.2 24B Instruct 2506?

Yes, at Q6_K (20.5 GB) or lower. Higher quantizations like Q8_0 (24.7 GB) exceed the NVIDIA GeForce RTX 4090's 24 GB.

What's the best quantization for Mistral Small 3.2 24B Instruct 2506?

For Mistral Small 3.2 24B Instruct 2506, Q4_K_M (15.1 GB) offers the best balance of quality and VRAM usage. Q4_K_L (15.4 GB) provides better quality if you have the VRAM. The smallest option is IQ2_XXS at 7.3 GB.

VRAM requirement by quantization

IQ2_XXS
7.3 GB
Q2_K
10.9 GB
Q3_K_L
13.0 GB
Q4_K_M ★
15.1 GB
Q5_0
15.7 GB
BF16
48.7 GB

★ Recommended — best balance of quality and VRAM usage.

Learn more about quantization →

Can I run Mistral Small 3.2 24B Instruct 2506 on a Mac?

Yes — Mac Mini M2 Pro (16 GB) and 133 other Macs can run Mistral Small 3.2 24B Instruct 2506, the smallest at IQ2_M. Apple Silicon uses unified memory, so the model shares RAM with the system. At Q4_K_M you need at least 15.1 GB of usable unified memory (RAM minus macOS overhead), from Mac Mini M5 Pro (24 GB).

Can I run Mistral Small 3.2 24B Instruct 2506 locally?

Yes — Mistral Small 3.2 24B Instruct 2506 can run locally on consumer hardware. Popular tools include Ollama, LM Studio, and llama.cpp.

How fast is Mistral Small 3.2 24B Instruct 2506?

At Q4_K_M, Mistral Small 3.2 24B Instruct 2506 can reach ~317 tok/s on AMD Instinct MI350X. On NVIDIA GeForce RTX 4090: ~43 tok/s. Speed depends mainly on GPU memory bandwidth. Real-world results typically within ±20%.

tok/s = (bandwidth GB/s ÷ model GB) × efficiency

Example: NVIDIA B200 → 8000 ÷ 15.1 × 0.65 = ~344 tok/s

Estimated speed at Q4_K_M (15.1 GB)

~344 tok/s
~43 tok/s
~344 tok/s
~317 tok/s

Real-world results typically within ±20%. Speed depends on batch size, quantization kernel, and software stack.

Learn more about tok/s estimation →

What's the download size of Mistral Small 3.2 24B Instruct 2506?

At Q4_K_M, the download is about 14.41 GB. The full-precision BF16 version is 48.02 GB. The smallest option (IQ2_XXS) is 6.60 GB.

Which GPUs can run Mistral Small 3.2 24B Instruct 2506?

26 consumer GPUs can run Mistral Small 3.2 24B Instruct 2506 at Q4_K_M. Top options include AMD Radeon RX 7900 XTX, NVIDIA GeForce RTX 3090, NVIDIA GeForce RTX 3090 Ti, AMD Radeon RX 6800. 7 GPUs have plenty of headroom for comfortable inference.

Which devices can run Mistral Small 3.2 24B Instruct 2506?

122 devices with unified memory can run Mistral Small 3.2 24B Instruct 2506 at Q4_K_M (15.1 GB), including AMD Ryzen AI 9 HX 370 (Strix Point) Laptop, ASUS Ascent GX10, Beelink GTR9 Pro (Ryzen AI Max+ 395, 128 GB), Framework Desktop (Ryzen AI Max+ 395, 128 GB). Apple Silicon Macs use unified memory shared between CPU and GPU, making them well-suited for local LLM inference.