Inkling-Small needs about 168.7 GB of memory at Q4_K_M (a 162.5 GB file) with an 8K context window, which is more than a 48 GB GPU holds. This repository has 23 quants, from 74.8 GB (IQ1_S) to 527.4 GB (BF16). Longer context needs more: at 32K tokens the KV cache alone adds about 22.5 GB.
Inkling‑Small: a multimodal 276B‑parameter model for text, image and audio
Inkling‑Small is a 42‑layer decoder‑only transformer with a sparse Mixture‑of‑Experts backbone. It accepts text, image and audio inputs and generates UTF‑8 text output. The model is open‑weight and designed for developers building conversational, instruction‑following and multimodal applications.
Good at
- Multimodal inference
- Large context up to 1,048,576 tokens
- Open‑weight research
Limits
- Requires 8‑k context for memory estimates
- No vision support in quantized versions
- License is Apache‑2.0
Best for
- Chatbots with image and audio input
- Coding assistants with multimodal prompts
- Retrieval‑augmented generation systems
Who it’s for. Developers building AI‑powered applications that need text, image and audio understanding. The model supports local deployment via SGLang, vLLM, TokenSpeed, Unsloth and Huggingface.
Quants & memory
| Quant | File | Est. memory | vs VRAM | Result | |
|---|---|---|---|---|---|
| IQ1_S | 74.8 GB (3 parts) | 80.9 GB | Too big | Get file → | |
| IQ1_M | 78.8 GB (3 parts) | 85.0 GB | Too big | Get file → | |
| IQ2_XXS | 82.3 GB (3 parts) | 88.4 GB | Too big | Get file → | |
| IQ2_M | 82.4 GB (3 parts) | 88.6 GB | Too big | Get file → | |
| Q2_K_XL | 87.9 GB (3 parts) | 94.1 GB | Too big | Get file → | |
| IQ3_XXS | 97.9 GB (3 parts) | 104.1 GB | Too big | Get file → | |
| IQ3_S | 107.5 GB (4 parts) | 113.6 GB | Too big | Get file → | |
| Q3_K_M | 119.4 GB (4 parts) | 125.5 GB | Too big | Get file → | |
| Q3_K_XL | 119.6 GB (4 parts) | 125.7 GB | Too big | Get file → | |
| IQ4_XS | 127.4 GB (4 parts) | 133.5 GB | Too big | Get file → | |
| IQ4_NL | 130.1 GB (4 parts) | 136.2 GB | Too big | Get file → | |
| Q4_K_S | 152.3 GB (5 parts) | 158.5 GB | Too big | Get file → | |
| MXFP4_MOE | 158.0 GB (5 parts) | 164.2 GB | Too big | Get file → | |
| Q4_K_M | 162.5 GB (5 parts) | 168.7 GB | Too big | Get file → | |
| Q4_K_XL | 163.3 GB (5 parts) | 169.4 GB | Too big | Get file → | |
| Q5_K_S | 184.2 GB (5 parts) | 190.3 GB | Too big | Get file → | |
| Q5_K_M | 196.1 GB (5 parts) | 202.2 GB | Too big | Get file → | |
| Q5_K_XL | 196.7 GB (5 parts) | 202.8 GB | Too big | Get file → | |
| Q6_K | 218.9 GB (6 parts) | 225.0 GB | Too big | Get file → | |
| Q6_K_XL | 239.7 GB (6 parts) | 245.8 GB | Too big | Get file → | |
| Q8_0 | 280.3 GB (7 parts) | 286.4 GB | Too big | Get file → | |
| Q8_K_XL | 288.4 GB (7 parts) | 294.6 GB | Too big | Get file → | |
| BF16 | 527.4 GB (11 parts) | 533.6 GB | Too big | Get file → |
For 8‑GB VRAM use IQ1_S (74.77 GB file, ~80.9 GB memory at 8k context). For 16‑GB VRAM use Q4_K_M (162.54 GB file, ~168.7 GB memory). For 32‑GB VRAM use Q5_K_XL (196.66 GB file, ~202.8 GB memory). Estimates are approximate.
What GPU do you need for Inkling-Small?
These figures add the file size, an fp16 KV cache at an 8K context window (worked out from this model’s own header) and about 0.5 GB for the runtime. They are estimates, so leave a little headroom, and expect to use less context or a smaller quant if you are right at the limit.
- 4 GB VRAM: no quant fits at an 8K context.
- 6 GB VRAM: no quant fits at an 8K context.
- 8 GB VRAM: no quant fits at an 8K context.
- 12 GB VRAM: no quant fits at an 8K context.
- 16 GB VRAM: no quant fits at an 8K context.
- 24 GB VRAM: no quant fits at an 8K context.
- 32 GB VRAM: no quant fits at an 8K context.
- 48 GB VRAM: no quant fits at an 8K context.
Which Inkling-Small quant should you choose?
Q4_K_M is the usual starting point for Inkling-Small: it balances file size and quality well. Here it is a 162.5 GB file.
The Q8_0 file is 280.3 GB, about 1.7 times the size of Q4_K_M. It keeps the weights closer to the original, which only matters if you have the memory to spare.
If memory is tight, the smallest file is IQ1_S at 74.8 GB, with the quality trade-offs that come with fewer bits. Use it only when memory is very tight. Expect a clear drop in quality, which is usually worse on small models and easier to live with on large ones.
New to quant names? Read GGUF quantization explained.
Specifications
Architecture from the GGUF header
Source
Frequently asked questions about Inkling-Small
How much VRAM does Inkling-Small need?
About 168.7 GB at Q4_K_M with an 8K context window. That is the 162.5 GB file, plus the KV cache and roughly 0.5 GB of runtime overhead. Smaller quants need less and longer context needs more.
Can I run Inkling-Small on an 8 GB GPU?
Not at an 8K context. The smallest file needs about 80.9 GB, so you would need a larger GPU or CPU offloading.
What is the file size of Inkling-Small GGUF?
It depends on the quant: IQ1_S is 74.8 GB, IQ1_M is 78.8 GB, IQ2_XXS is 82.3 GB, IQ2_M is 82.4 GB, Q2_K_XL is 87.9 GB, IQ3_XXS is 97.9 GB and 17 more.
What context length does Inkling-Small support?
The GGUF header lists a trained context length of 1,048,576 tokens (1024K). The KV cache grows with the context you use: about 5.6 GB at 8K and 22.5 GB at 32K.
Where can I download Inkling-Small GGUF?
The files are on Hugging Face in the lyssquant/Inkling-Small-GGUF repository. The download buttons on this page open the file on Hugging Face. ggufmodels.com does not host any model files.
What license does Inkling-Small use?
Hugging Face lists the license as apache-2.0. Licenses can differ between the quantized files and the original model, so check the repository and the base model before commercial use.
Can I run Inkling‑Small on 8 GB VRAM?
The smallest quantized file is IQ1_S, which requires about 80.9 GB of memory at 8k context, so 8 GB VRAM is insufficient.
Does Inkling‑Small support vision?
Yes, it accepts image inputs via a hierarchical patch encoder.
Is the model licensed for commercial use?
It is released under Apache‑2.0, which permits commercial use with attribution.