CohereLabs · text

c4ai-command-r-plus-08-2024

CohereLabs/c4ai-command-r-plus-08-2024

c4ai-command-r-plus-08-2024 at Q4_K_M is exactly 62,750,608,128 bytes (58.44 GiB / 62.75 GB) — an effective 4.836 bits per weight, not the nominal 4. Its KV cache at 32K is 8.00 GiB.

From the file· summed from 2 file(s)From the file· KV from mirror (mirror:unsloth/c4ai-command-r-plus-08-2024)
Parameters
104B
Architecture
command-r
64 layers
Context
131,072
native (config.json)
License
cc-by-nc-4.0

Shipped quantizations

● exact bytes, summed from published files
QuantSize●Exact bytes●Effective bpw●Tensors●Publisher
IQ1_S21.59 GiB23,181,871,9361.786—legraphista
IQ1_M23.49 GiB25,217,944,3841.943—legraphista
IQ1_M23.49 GiB25,217,944,3841.943—bartowski
IQ2_XXS26.65 GiB28,611,398,4642.205—bartowski
IQ2_XXS26.65 GiB28,611,398,4642.205—legraphista
IQ2_XS29.46 GiB31,628,151,6162.437—bartowski
IQ2_XS29.46 GiB31,628,151,6162.437—legraphista
IQ2_S31.04 GiB33,324,485,4402.568—legraphista
IQ2_M33.56 GiB36,039,248,7042.777—legraphista
IQ2_M33.56 GiB36,039,248,7042.777—bartowski
Q2_K_S34.08 GiB36,595,452,7362.820—legraphista
Q2_K36.78 GiB39,497,386,8163.044—legraphista
Q2_K36.78 GiB39,497,386,8163.044—bartowski
Q2_K_L37.49 GiB40,259,242,8163.103—bartowski
IQ3_XXS37.87 GiB40,658,750,2723.133—legraphista
IQ3_XXS37.87 GiB40,658,750,2723.133—bartowski
IQ3_XS40.61 GiB43,599,416,1283.360—legraphista
Q3_K_S42.70 GiB45,851,757,3763.534—bartowski
Q3_K_S2 shards42.70 GiB45,851,757,5683.534—legraphista
IQ3_S2 shards42.80 GiB45,958,712,3203.542—legraphista
IQ3_M44.41 GiB47,683,357,5043.675—bartowski
IQ3_M2 shards44.41 GiB47,683,357,6963.675—legraphista
Q3_K_M2 shards47.48 GiB50,982,439,9363.929—bartowski
Q3_K3 shards47.48 GiB50,982,440,0643.929—legraphista
Q3_K_L2 shards51.60 GiB55,402,187,4884.269—lmstudio-community
Q3_K_L2 shards51.60 GiB55,402,187,7764.269—bartowski
Q3_K_L3 shards51.60 GiB55,402,187,9044.269—legraphista
IQ4_XS2 shards52.34 GiB56,201,202,7204.331—bartowski
IQ4_XS3 shards52.34 GiB56,201,202,8164.331—legraphista
IQ4_NL3 shards55.25 GiB59,321,764,9924.572—legraphista
Q4_02 shards55.35 GiB59,428,719,6484.580—bartowski
Q4_K_S2 shards55.55 GiB59,642,629,1204.596—bartowski
Q4_K_S3 shards55.55 GiB59,642,629,2484.596—legraphista
Q4_K_M2 shards58.44 GiB62,750,608,1284.836—lmstudio-community
Q4_K_M2 shards58.44 GiB62,750,608,4164.836—bartowski
Q4_K3 shards58.44 GiB62,750,608,5124.836—legraphista
Q4_K_L2 shards59.15 GiB63,512,464,3844.894—bartowski
Q5_K_S3 shards66.87 GiB71,804,013,4085.534—legraphista
Q5_K4 shards68.57 GiB73,622,244,3205.674—legraphista
Q5_K_M2 shards68.57 GiB73,622,244,3845.674—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-E4B-itvs Qwen3-30B-A3B-Thinking-2507vs Qwen3-8BAll 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 54.38 GiB. The real file is 58.44 GiB, because a quantization is a mixture and some tensors are always kept at higher precision.

Architecture

from mirror:unsloth/c4ai-command-r-plus-08-2024
Layers
64
Attention heads
96
KV heads
8
Head dim
128
Hidden size
12288
Vocab
256,000
Sliding window
none
SWA period
—
MLA
no
Experts
—
Experts per token
—
use_sliding_window
—

Questions people ask

How much VRAM does c4ai-command-r-plus-08-2024 need?
Q4_K_M is exactly 62,750,608,128 bytes (58.44 GiB) in weights. Add the KV cache, which depends on your context length, plus roughly half a gigabyte of runtime overhead.
How large is c4ai-command-r-plus-08-2024'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 c4ai-command-r-plus-08-2024 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.