Inkling-Small GGUF size and VRAM requirements

License: apache-2.0 ⬇ 1,172,415 ❤ 83
Parameters265.96B
Context1,048,576

unsloth/Inkling-Small-GGUF is a very large language model with 265.96 billion parameters, built on the inkling architecture. It is released under the apache-2.0 license and has been downloaded 1,172,415 times.

Inkling-Small-GGUF is a Mixture-of-Experts model with 256 experts, of which 6 are active on each token. Routing only 6 of 256 experts makes it noticeably faster than a dense model of the same size — but every expert still has to be held in memory, so the numbers below are set by the full parameter count, not the active one. It is built from 42 transformer layers, a hidden size of 4,096, 32 attention heads. It uses grouped-query attention (32 query heads sharing 8 key/value heads), which already trims KV-cache memory compared with full multi-head attention.

To run unsloth/Inkling-Small-GGUF locally at a 4,096-token context, its quantized versions need between 71.09 GB (IQ1_S, lowest quality) and 492.65 GB (BF16, highest quality) of memory, weights plus KV cache and a system margin included. Context length drives that memory directly: at 4,096 tokens the KV cache for Inkling-Small-GGUF is about 0.66 GB, rising to roughly 168.0 GB at its full 1,048,576-token context. Shorter prompts free up memory for a higher-quality quantization.

Available GGUF quantizations for unsloth/Inkling-Small-GGUF include IQ1_S, IQ1_M, IQ2_XXS, IQ2_M, Q2_K_XL, IQ3_XXS, IQ3_S, Q3_K_M, Q3_K_XL, IQ4_XS, IQ4_NL, Q4_K_S, MXFP4, Q4_K_M, Q4_K_XL, Q5_K_S, Q5_K_M, Q5_K_XL, Q6_K, Q6_K_XL, Q8_0, Q8_K_XL, BF16. The model supports a native context length of up to 1,048,576 tokens; a longer context grows the KV cache and the memory needed.

→ Guide: How much VRAM do you need?

GGUF file size and memory by quantization

Compare real GGUF weight sizes, estimated KV cache and total memory for Q4, Q5, Q8 and every quantization published in this repository.

Quant.Bits QualityWeights KVTotal Speed~Verdict
IQ1_S 2.25 Very low 69.64 GB 0.66 GB 71.09 GB Insufficient
IQ1_M 2.37 Very low 73.42 GB 0.66 GB 74.87 GB Insufficient
IQ2_XXS 2.48 Very low 76.64 GB 0.66 GB 78.09 GB Insufficient
IQ2_M 2.48 Very low 76.78 GB 0.66 GB 78.23 GB Insufficient
Q2_K_XL 2.65 Low 81.9 GB 0.66 GB 83.36 GB Insufficient
IQ3_XXS 2.95 Low 91.21 GB 0.66 GB 92.66 GB Insufficient
IQ3_S 3.23 Low 100.08 GB 0.66 GB 101.54 GB Insufficient
Q3_K_M 3.59 Fair 111.16 GB 0.66 GB 112.61 GB Insufficient
Q3_K_XL 3.6 Fair 111.34 GB 0.66 GB 112.8 GB Insufficient
IQ4_XS 3.83 Fair 118.66 GB 0.66 GB 120.11 GB Insufficient
IQ4_NL 3.91 Fair 121.16 GB 0.66 GB 122.61 GB Insufficient
Q4_K_S 4.58 Good 141.87 GB 0.66 GB 143.33 GB Insufficient
MXFP4 4.75 Good 147.19 GB 0.66 GB 148.64 GB Insufficient
Q4_K_M 4.89 Good 151.37 GB 0.66 GB 152.83 GB Insufficient
Q4_K_XL 4.91 Good 152.06 GB 0.66 GB 153.52 GB Insufficient
Q5_K_S 5.54 Very good 171.56 GB 0.66 GB 173.02 GB Insufficient
Q5_K_M 5.9 Very good 182.62 GB 0.66 GB 184.08 GB Insufficient
Q5_K_XL 5.92 Very good 183.15 GB 0.66 GB 184.61 GB Insufficient
Q6_K 6.58 Excellent 203.87 GB 0.66 GB 205.33 GB Insufficient
Q6_K_XL 7.21 Excellent 223.25 GB 0.66 GB 224.7 GB Insufficient
Q8_0 8.43 Excellent 261.03 GB 0.66 GB 262.49 GB Insufficient
Q8_K_XL 8.68 Excellent 268.61 GB 0.66 GB 270.07 GB Insufficient
BF16 15.86 Excellent 491.19 GB 0.66 GB 492.65 GB Insufficient

KV cache computed from the model's exact architecture. Speed is a rough estimate bounded by memory bandwidth.

Frequently asked questions

What kind of model is unsloth/Inkling-Small-GGUF?

unsloth/Inkling-Small-GGUF is a language model with 265.96 billion parameters, based on the inkling architecture. It is released under the apache-2.0 license and distributed as GGUF files for local inference.

Is unsloth/Inkling-Small-GGUF a Mixture-of-Experts (MoE) model?

Yes. unsloth/Inkling-Small-GGUF is a Mixture-of-Experts model with 256 experts, of which 6 are activated per token. That makes it faster than a dense model of the same size, but all 256 experts must be loaded into memory, so the VRAM/RAM it needs is driven by the total parameter count, not the active one.

Can I run unsloth/Inkling-Small-GGUF on an 8 GB GPU?

No. unsloth/Inkling-Small-GGUF does not fit on an 8 GB GPU, even with the smallest quantization and system RAM offloading.

Can I run unsloth/Inkling-Small-GGUF on a 16 GB GPU?

No. unsloth/Inkling-Small-GGUF does not fit on a 16 GB GPU, even with the smallest quantization and system RAM offloading.

Can I run unsloth/Inkling-Small-GGUF on a 24 GB GPU?

Partially. unsloth/Inkling-Small-GGUF only fits on a 24 GB GPU by offloading part of it to system RAM (with IQ1_S), which runs but is slower.

What context length does unsloth/Inkling-Small-GGUF support?

unsloth/Inkling-Small-GGUF supports a native context length of up to 1,048,576 tokens. A longer context grows the KV cache, so it increases the memory needed to run the model.

What is the best quantization for unsloth/Inkling-Small-GGUF?

For unsloth/Inkling-Small-GGUF, a strong default is Q4_K_M, which needs about 152.83 GB and keeps most of the quality while roughly halving the memory versus 8-bit. With VRAM to spare, Q5_K_M or Q6_K add a little more quality; if you are tight on memory, a smaller quantization still runs. Pick the highest quantization that fits your VRAM.