ByteDance-Seed · text

Seed-OSS-36B-Instruct

ByteDance-Seed/Seed-OSS-36B-Instruct

Seed-OSS-36B-Instruct at Q4_K_M is exactly 21,762,148,864 bytes (20.27 GiB / 21.76 GB) — an effective 4.816 bits per weight, not the nominal 4. Its KV cache at 32K is 8.00 GiB.

From the file· summed from 1 file(s)From the file· KV per layer
Parameters
36.2B
Architecture
seed_oss
64 layers
Context
524,288
native (config.json)
License
apache-2.0

Shipped quantizations

● exact bytes, summed from published files
QuantSize●Exact bytes●Effective bpw●Tensors●Publisher
UD-IQ1_S7.89 GiB8,473,742,4961.875—unsloth
UD-IQ1_M8.46 GiB9,080,892,5762.010—unsloth
IQ2_XXS9.23 GiB9,910,925,9842.193—bartowski
UD-IQ2_XXS9.46 GiB10,160,802,9762.248—unsloth
IQ2_XS10.19 GiB10,945,084,0642.422—bartowski
IQ2_S10.82 GiB11,612,629,6642.570—bartowski
IQ2_M11.68 GiB12,541,930,1442.775—bartowski
UD-IQ2_M11.86 GiB12,729,568,4162.817—unsloth
Q2_K12.67 GiB13,604,186,7843.010—bartowski
Q2_K12.67 GiB13,604,187,2963.010—unsloth
Q2_K_L12.84 GiB13,790,350,4963.052—unsloth
IQ3_XXS13.15 GiB14,116,760,2243.124—bartowski
UD-IQ3_XXS13.27 GiB14,253,649,0563.154—unsloth
Q2_K_L13.39 GiB14,379,866,7843.182—bartowski
IQ3_XS14.05 GiB15,089,949,3443.339—bartowski
Q3_K_S14.77 GiB15,855,409,8243.509—bartowski
Q3_K_S14.77 GiB15,855,410,3363.509—unsloth
IQ3_M15.36 GiB16,496,024,2243.651—bartowski
Q3_K_M16.41 GiB17,620,949,6643.899—bartowski
Q3_K_M16.41 GiB17,620,950,1763.899—unsloth
Q3_K_L17.83 GiB19,142,695,5844.236—bartowski
IQ4_XS18.16 GiB19,498,617,5044.315—bartowski
IQ4_XS18.18 GiB19,523,439,7764.320—unsloth
IQ4_NL19.18 GiB20,592,986,7844.557—bartowski
IQ4_NL19.18 GiB20,592,987,2964.557—unsloth
Q4_019.21 GiB20,621,822,6244.564—bartowski
Q4_019.21 GiB20,621,823,1364.564—unsloth
Q4_K_S19.27 GiB20,695,222,9444.580—bartowski
Q4_K_S19.27 GiB20,695,223,4564.580—unsloth
Q4_K_M20.27 GiB21,762,148,8644.816—lmstudio-community
Q4_K_M20.27 GiB21,762,149,0244.816—bartowski
Q4_K_M20.27 GiB21,762,149,5364.816—unsloth
Q4_K_L20.82 GiB22,351,665,8244.946—bartowski
Q4_121.20 GiB22,760,753,8245.037—bartowski
Q4_121.20 GiB22,760,754,3365.037—unsloth
Q5_K_S23.26 GiB24,970,463,9045.526—bartowski
Q5_K_S23.26 GiB24,970,464,4165.526—unsloth
Q5_K_M23.84 GiB25,594,366,6245.664—bartowski
Q5_K_M23.84 GiB25,594,367,1365.664—unsloth
Q5_K_L24.29 GiB26,084,596,3845.772—bartowski

KV cache by context

computed per layer
ContextKV cache (f16)●Flat formulaOverstated byFull / windowed / recurrent
4,0961.00 GiB1.00 GiB—64 / 0 / 0
8,1922.00 GiB2.00 GiB—64 / 0 / 0
16,3844.00 GiB4.00 GiB—64 / 0 / 0
32,7688.00 GiB8.00 GiB—64 / 0 / 0
65,53616.00 GiB16.00 GiB—64 / 0 / 0
131,07232.00 GiB32.00 GiB—64 / 0 / 0

Compare with

same modality, comparable size
vs Qwen3-Coder-30B-A3B-Instructvs Qwen3.6-27Bvs Qwen3.8-27Bvs gemma-4-12B-it-qat-q4_0-unquantizedvs gemma-4-E4B-itvs Qwen3-30B-A3B-Thinking-2507All 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 18.94 GiB. The real file is 20.27 GiB, because a quantization is a mixture and some tensors are always kept at higher precision.

Architecture

from config.json
Layers
64
Attention heads
80
KV heads
8
Head dim
128
Hidden size
5120
Vocab
155,136
Sliding window
none
SWA period
—
MLA
no
Experts
—
Experts per token
—
use_sliding_window
—

Questions people ask

How much VRAM does Seed-OSS-36B-Instruct need?
Q4_K_M is exactly 21,762,148,864 bytes (20.27 GiB) in weights. Add the KV cache, which depends on your context length, plus roughly half a gigabyte of runtime overhead.
How large is Seed-OSS-36B-Instruct's KV cache?
8.00 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 Seed-OSS-36B-Instruct 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.