CohereLabs · text

c4ai-command-r-v01

CohereLabs/c4ai-command-r-v01

c4ai-command-r-v01 at Q4_K_M is exactly 21,527,051,232 bytes (20.05 GiB / 21.53 GB) — an effective 4.923 bits per weight, not the nominal 4.

From the file· summed from 1 file(s)
Parameters
35.0B
Architecture
command-r
Context
—
native (config.json)
License
apache-2.0

Shipped quantizations

● exact bytes, summed from published files
QuantSize●Exact bytes●Effective bpw●Tensors●Publisher
IQ1_S7.94 GiB8,523,398,1121.949—dranger003
IQ1_M8.52 GiB9,146,645,4722.092—dranger003
IQ2_XXS9.49 GiB10,185,391,0722.329—dranger003
IQ2_XS10.34 GiB11,100,273,6322.539—dranger003
IQ2_S11.03 GiB11,844,107,2322.709—dranger003
IQ2_M11.80 GiB12,675,103,7122.899—dranger003
Q2_K_S11.86 GiB12,738,673,6322.913—dranger003
Q2_K12.87 GiB13,817,395,8723.160—DavidAU
Q2_K12.87 GiB13,817,396,1923.160—dranger003
IQ3_XXS12.88 GiB13,832,469,4723.163—dranger003
IQ3_XS14.05 GiB15,091,416,0323.451—dranger003
IQ3_S14.77 GiB15,862,119,3923.628—dranger003
Q3_K_S14.77 GiB15,862,119,3923.628—dranger003
IQ3_M15.55 GiB16,697,703,3923.819—dranger003
Q3_K_M16.41 GiB17,618,484,1924.029—dranger003
Q3_K_L17.83 GiB19,149,405,1524.379—dranger003
IQ4_XS17.88 GiB19,201,833,9524.391—dranger003
IQ4_NL18.84 GiB20,229,438,4324.626—dranger003
Q4_K_S18.98 GiB20,378,336,2244.660—dranger003
Q4_K_M20.05 GiB21,527,051,2324.923—dranger003
IQ2_XS2 shards20.68 GiB22,200,547,2005.077—DavidAU
Q5_K_S22.67 GiB24,339,856,3525.566—dranger003
Q5_K_M23.29 GiB25,008,323,5525.719—dranger003
Q6_K26.74 GiB28,707,175,1046.565—dranger003
IQ3_XS2 shards28.11 GiB30,182,832,0006.903—DavidAU
IQ3_M2 shards31.10 GiB33,395,406,7207.637—DavidAU
Q8_034.63 GiB37,179,013,8248.503—dranger003
Q4_K_S2 shards37.96 GiB40,756,672,3849.321—DavidAU
Q5_K_S2 shards45.34 GiB48,679,712,64011.133—DavidAU
Q6_K2 shards53.47 GiB57,414,350,72013.130—DavidAU
F162 shards65.17 GiB69,973,228,38416.003—dranger003

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 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 18.33 GiB. The real file is 20.05 GiB, because a quantization is a mixture and some tensors are always kept at higher precision.

Architecture

Architecture unavailable — this repository is gated and no ungated mirror was found. Exact file sizes above are still authoritative; only the KV math needs the config.

Questions people ask

How much VRAM does c4ai-command-r-v01 need?
Q4_K_M is exactly 21,527,051,232 bytes (20.05 GiB) in weights. Add the KV cache, which depends on your context length, plus roughly half a gigabyte of runtime overhead.
Which quantization of c4ai-command-r-v01 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.