cerebras · text · mixture of experts

Qwen3-Coder-REAP-363B-A35B

cerebras/Qwen3-Coder-REAP-363B-A35B

Qwen3-Coder-REAP-363B-A35B at Q4_K_M is exactly 219,307,470,592 bytes (204.25 GiB / 219.31 GB) — an effective 4.832 bits per weight, not the nominal 4. Its KV cache at 32K is 7.75 GiB.

From the file· summed from 5 file(s)From the file· KV per layer
Parameters
363B
total, not active
Architecture
qwen3moe
62 layers
Context
262,144
native (config.json)
License
apache-2.0

Shipped quantizations

● exact bytes, summed from published files
QuantSize●Exact bytes●Effective bpw●Tensors●Publisher
UD-TQ1_094.37 GiB101,329,153,2162.232—unsloth
UD-IQ1_S3 shards97.24 GiB104,410,983,9362.300—unsloth
UD-IQ1_M3 shards107.85 GiB115,803,287,0722.551—unsloth
Q2_K3 shards123.15 GiB132,236,213,7602.913—unsloth
Q2_K_L3 shards123.36 GiB132,455,001,6322.918—unsloth
UD-IQ3_XXS4 shards144.19 GiB154,820,144,7683.411—unsloth
Q3_K_S4 shards145.81 GiB156,564,137,6323.449—unsloth
Q3_K_M4 shards161.57 GiB173,488,842,4003.822—unsloth
IQ4_XS4 shards183.56 GiB197,090,864,7684.342—unsloth
IQ4_NL5 shards190.65 GiB204,709,424,9284.510—unsloth
Q4_05 shards191.37 GiB205,486,419,7444.527—unsloth
Q4_K_S5 shards192.15 GiB206,315,319,0724.545—unsloth
Q4_K_M5 shards204.25 GiB219,307,470,5924.832—unsloth
Q4_15 shards211.68 GiB227,294,326,5605.008—unsloth
Q5_K_S6 shards232.76 GiB249,927,987,0725.506—unsloth
Q5_K_M6 shards239.79 GiB257,473,310,6245.673—unsloth
Q6_K7 shards277.56 GiB298,024,515,5846.566—unsloth
Q8_08 shards359.45 GiB385,956,030,2728.503—unsloth
BF1616 shards676.45 GiB726,336,088,70416.002—unsloth

KV cache by context

computed per layer
ContextKV cache (f16)●Flat formulaOverstated byFull / windowed / recurrent
4,0960.97 GiB0.97 GiB—62 / 0 / 0
8,1921.94 GiB1.94 GiB—62 / 0 / 0
16,3843.88 GiB3.88 GiB—62 / 0 / 0
32,7687.75 GiB7.75 GiB—62 / 0 / 0
65,53615.50 GiB15.50 GiB—62 / 0 / 0
131,07231.00 GiB31.00 GiB—62 / 0 / 0

Compare with

same modality, comparable size
vs Qwen3-Coder-30B-A3B-Instructvs Qwen3.6-27Bvs Qwen3.8-27Bvs DeepSeek-V4-Flashvs Qwen3-30B-A3B-Thinking-2507vs Laguna-XS-2.1All alternatives →Will it fit on my card? →

Will it run on your card?

full quant x context sweep
By VRAM:8GB12GB16GB24GB32GB48GB
Radeon RX 6500 XT 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 D 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 190.23 GiB. The real file is 204.25 GiB, because a quantization is a mixture and some tensors are always kept at higher precision.

Architecture

from config.json
Layers
62
Attention heads
96
KV heads
8
Head dim
128
Hidden size
6144
Vocab
151,936
Sliding window
none
SWA period
—
MLA
no
Experts
120
Experts per token
8
use_sliding_window
false

Questions people ask

How much VRAM does Qwen3-Coder-REAP-363B-A35B need?
Q4_K_M is exactly 219,307,470,592 bytes (204.25 GiB) in weights. Add the KV cache, which depends on your context length, plus roughly half a gigabyte of runtime overhead.
How large is Qwen3-Coder-REAP-363B-A35B's KV cache?
7.75 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.
Is Qwen3-Coder-REAP-363B-A35B a mixture-of-experts model?
Yes — 120 experts, 8 routed per token. Every expert must be resident, but only the routed ones are read per token, which is why its memory requirement and its speed behave very differently.
Which quantization of Qwen3-Coder-REAP-363B-A35B 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.