mistralai · text

Mathstral-7B-v0.1

mistralai/Mathstral-7B-v0.1

Mathstral-7B-v0.1 at Q4_K_M is exactly 4,372,811,584 bytes (4.07 GiB / 4.37 GB) — an effective 4.827 bits per weight, not the nominal 4. Its KV cache at 32K is 4.00 GiB.

From the file· summed from 1 file(s)From the file· KV per layer
Parameters
7.2B
Architecture
llama
32 layers
Context
32,768
native (config.json)
License
—

Shipped quantizations

● exact bytes, summed from published files
QuantSize●Exact bytes●Effective bpw●Tensors●Publisher
IQ1_S1.50 GiB1,615,318,8481.783—MaziyarPanahi
IQ1_M1.64 GiB1,757,663,0401.940—MaziyarPanahi
IQ1_M1.64 GiB1,757,663,0721.940—InferenceIllusionist
IQ2_XXS1.86 GiB1,994,903,3922.202—InferenceIllusionist
IQ2_XS2.05 GiB2,201,472,8322.430—MaziyarPanahi
IQ2_XS2.05 GiB2,201,472,8642.430—InferenceIllusionist
IQ2_S2.16 GiB2,314,456,9282.555—InferenceIllusionist
IQ2_M2.33 GiB2,504,249,1522.764—bartowski
IQ2_M2.33 GiB2,504,249,1842.764—InferenceIllusionist
Q2_K2.54 GiB2,722,877,2483.005—MaziyarPanahi
Q2_K2.54 GiB2,722,877,2483.005—bartowski
Q2_K2.54 GiB2,722,877,2803.005—InferenceIllusionist
IQ3_XXS2.64 GiB2,830,880,6083.125—InferenceIllusionist
Q2_K_L2.66 GiB2,853,949,2483.150—bartowski
IQ3_XS2.82 GiB3,022,769,9843.336—bartowski
IQ3_XS2.82 GiB3,022,769,9843.336—MaziyarPanahi
IQ3_XS2.82 GiB3,022,770,0163.336—InferenceIllusionist
Q3_K_S2.95 GiB3,168,522,0483.497—bartowski
Q3_K_S2.95 GiB3,168,522,0483.497—MaziyarPanahi
Q3_K_S2.95 GiB3,168,522,0803.497—InferenceIllusionist
IQ3_S2.97 GiB3,186,347,8723.517—InferenceIllusionist
IQ3_M3.06 GiB3,288,846,1443.630—bartowski
IQ3_M3.06 GiB3,288,846,1763.630—InferenceIllusionist
Q3_K_M3.28 GiB3,522,940,7363.888—bartowski
Q3_K_M3.28 GiB3,522,940,7363.888—MaziyarPanahi
Q3_K_M3.28 GiB3,522,940,7683.888—InferenceIllusionist
Q3_K_L3.56 GiB3,825,979,2004.223—bartowski
Q3_K_L3.56 GiB3,825,979,2004.223—lmstudio-community
Q3_K_L3.56 GiB3,825,979,2004.223—MaziyarPanahi
IQ4_XS3.64 GiB3,911,962,4324.318—bartowski
IQ4_XS3.64 GiB3,911,962,4324.318291MaziyarPanahi
IQ4_XS3.64 GiB3,911,962,4324.318—lmstudio-community
IQ4_XS3.64 GiB3,911,962,4644.318—InferenceIllusionist
Q4_K_S3.86 GiB4,144,746,3044.575—MaziyarPanahi
Q4_K_S3.86 GiB4,144,746,3044.575—bartowski
Q4_K_S3.86 GiB4,144,746,3364.575—InferenceIllusionist
Q4_K_M4.07 GiB4,372,811,5844.827291MaziyarPanahi
Q4_K_M4.07 GiB4,372,811,5844.827—bartowski
Q4_K_M4.07 GiB4,372,811,5844.827—lmstudio-community
Q4_K_M4.07 GiB4,372,811,6164.827—InferenceIllusionist

KV cache by context

computed per layer
ContextKV cache (f16)●Flat formulaOverstated byFull / windowed / recurrent
4,0960.50 GiB0.50 GiB—32 / 0 / 0
8,1921.00 GiB1.00 GiB—32 / 0 / 0
16,3842.00 GiB2.00 GiB—32 / 0 / 0
32,7684.00 GiB4.00 GiB—32 / 0 / 0
65,5368.00 GiB8.00 GiB—32 / 0 / 0
131,07216.00 GiB16.00 GiB—32 / 0 / 0

Compare with

same modality, comparable size
vs gemma-4-E4B-itvs Qwen3-4Bvs Qwen3-8Bvs HyperCLOVAX-SEED-Text-Instruct-1.5Bvs Llama-3.2-1B-Instructvs Qwen3-VL-8B-Instruct-abliterated-v1All alternatives →Will it fit on my card? →

Will it run on your card?

full quant x context sweep
By VRAM:8GB12GB16GB24GB32GB48GB
Radeon RX 6400 4GBGeForce RTX 3050 6GBGeForce RTX 5050 8GBGeForce RTX 3080 10GBGeForce RTX 2080 Ti 11GBGeForce RTX 5070 12GBGeForce RTX 5060 Ti 16GBApple M3 Pro 18GBGeForce RTX 3080 Ti 20GBGeForce RTX 5090 D V2 24GBGeForce RTX 5090 32GBApple M5 Max 36GBApple M5 Pro 48GBApple M5 Max 64GBApple M3 Ultra 96GBApple M5 Max 128GBApple M2 Ultra 192GBApple M3 Ultra 256GBApple M3 Ultra 512GB

Why other calculators give a different number

A parameters × bits ÷ 8 estimate puts Q4_K_M at roughly 3.80 GiB. The real file is 4.07 GiB, because a quantization is a mixture and some tensors are always kept at higher precision.

Architecture

from config.json
Layers
32
Attention heads
32
KV heads
8
Head dim
128
Hidden size
4096
Vocab
32,768
Sliding window
none
SWA period
—
MLA
no
Experts
—
Experts per token
—
use_sliding_window
—

Questions people ask

How much VRAM does Mathstral-7B-v0.1 need?
Q4_K_M is exactly 4,372,811,584 bytes (4.07 GiB) in weights. Add the KV cache, which depends on your context length, plus roughly half a gigabyte of runtime overhead.
How large is Mathstral-7B-v0.1's KV cache?
4.00 GiB at 32K context with an f16 cache, computed per layer. Quantizing the cache to q8_0 roughly halves it, which is often the difference between a context length fitting and not.
Which quantization of Mathstral-7B-v0.1 should I use?
Q4_K_M is the usual default. Pick the largest quantization that fits your card at the context you actually need — the table above gives exact sizes for every one published.