Iris 12B v1.4 needs about 20.8 GB of memory at Q4_K_M (a 7.4 GB file) with an 8K context window, so it fits on a 24 GB GPU. This repository has 25 quants, from 0.0 GB to 9.8 GB (Q6_K). Longer context needs more: at 32K tokens the KV cache alone adds about 51.5 GB.
Iris 12B v1.4 is a 11.9B-parameter Iris model in GGUF format, quantized by mradermacher. Its trained context length is 256K tokens. The repository has 25 quant files, from 0.0 GB to 9.8 GB (Q6_K).
Quants & memory
| Quant | File | Est. memory | vs VRAM | Result | |
|---|---|---|---|---|---|
| — | 0.0 GB | 13.4 GB | Too big | Get file → | |
| IQ1_S | 3.0 GB | 16.4 GB | Too big | Get file → | |
| IQ1_M | 3.2 GB | 16.6 GB | Too big | Get file → | |
| IQ2_XXS | 3.6 GB | 17.0 GB | Too big | Get file → | |
| IQ2_XS | 3.9 GB | 17.3 GB | Too big | Get file → | |
| IQ2_S | 4.1 GB | 17.5 GB | Too big | Get file → | |
| IQ2_M | 4.4 GB | 17.8 GB | Too big | Get file → | |
| Q2_K_S | 4.5 GB | 17.9 GB | Too big | Get file → | |
| Q2_K | 4.8 GB | 18.2 GB | Too big | Get file → | |
| IQ3_XXS | 4.8 GB | 18.2 GB | Too big | Get file → | |
| IQ3_XS | 5.3 GB | 18.7 GB | Too big | Get file → | |
| IQ3_S | 5.5 GB | 18.9 GB | Too big | Get file → | |
| Q3_K_S | 5.5 GB | 18.9 GB | Too big | Get file → | |
| IQ3_M | 5.7 GB | 19.1 GB | Too big | Get file → | |
| Q3_K_M | 6.1 GB | 19.5 GB | Too big | Get file → | |
| Q3_K_L | 6.6 GB | 20.0 GB | Too big | Get file → | |
| IQ4_XS | 6.6 GB | 20.0 GB | Too big | Get file → | |
| IQ4_NL | 7.0 GB | 20.4 GB | Too big | Get file → | |
| Q4_0 | 7.0 GB | 20.4 GB | Too big | Get file → | |
| Q4_K_S | 7.0 GB | 20.4 GB | Too big | Get file → | |
| Q4_K_M | 7.4 GB | 20.8 GB | Too big | Get file → | |
| Q4_1 | 7.7 GB | 21.0 GB | Too big | Get file → | |
| Q5_K_S | 8.3 GB | 21.7 GB | Too big | Get file → | |
| Q5_K_M | 8.5 GB | 21.9 GB | Too big | Get file → | |
| Q6_K | 9.8 GB | 23.2 GB | Too big | Get file → |
What GPU do you need for Iris 12B v1.4?
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: the best quant that fits is the main file (0.0 GB file, about 13.4 GB in use).
- 24 GB VRAM: the best quant that fits is Q6_K (9.8 GB file, about 23.2 GB in use).
Which Iris 12B v1.4 quant should you choose?
Q4_K_M is the usual starting point for Iris 12B v1.4: it balances file size and quality well. Here it is a 7.4 GB file.
If memory is tight, the smallest file is the first one in the table at 0.0 GB, with the quality trade-offs that come with fewer bits.
New to quant names? Read GGUF quantization explained.
Specifications
Architecture from the GGUF header
Source
Frequently asked questions about Iris 12B v1.4
How much VRAM does Iris 12B v1.4 need?
About 20.8 GB at Q4_K_M with an 8K context window. That is the 7.4 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 Iris 12B v1.4 on an 8 GB GPU?
Not at an 8K context. The smallest file needs about 13.4 GB, so you would need a larger GPU or CPU offloading.
What is the file size of Iris 12B v1.4 GGUF?
It depends on the quant: file is 0.0 GB, IQ1_S is 3.0 GB, IQ1_M is 3.2 GB, IQ2_XXS is 3.6 GB, IQ2_XS is 3.9 GB, IQ2_S is 4.1 GB and 19 more.
What context length does Iris 12B v1.4 support?
The GGUF header lists a trained context length of 262,144 tokens (256K). The KV cache grows with the context you use: about 12.9 GB at 8K and 51.5 GB at 32K.
Where can I download Iris 12B v1.4 GGUF?
The files are on Hugging Face in the mradermacher/Iris-12B-v1.4-i1-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 Iris 12B v1.4 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.