--- base_model: - Himitsui/Kaiju-11B - Sao10K/Fimbulvetr-11B-v2 - decapoda-research/Antares-11b-v2 - beberik/Nyxene-v3-11B exported_from: Steelskull/Umbra-v3-MoE-4x11b language: - en library_name: transformers license: apache-2.0 quantized_by: mradermacher tags: - moe - frankenmoe - merge - mergekit - Himitsui/Kaiju-11B - Sao10K/Fimbulvetr-11B-v2 - decapoda-research/Antares-11b-v2 - beberik/Nyxene-v3-11B --- ## About weighted/imatrix quants of https://huggingface.co/Steelskull/Umbra-v3-MoE-4x11b static quants are available at https://huggingface.co/mradermacher/Umbra-v3-MoE-4x11b-GGUF ## Usage If you are unsure how to use GGUF files, refer to one of [TheBloke's READMEs](https://huggingface.co/TheBloke/KafkaLM-70B-German-V0.1-GGUF) for more details, including on how to concatenate multi-part files. ## Provided Quants (sorted by size, not necessarily quality. IQ-quants are often preferable over similar sized non-IQ quants) | Link | Type | Size/GB | Notes | |:-----|:-----|--------:|:------| | [GGUF](https://huggingface.co/mradermacher/Umbra-v3-MoE-4x11b-i1-GGUF/resolve/main/Umbra-v3-MoE-4x11b.i1-IQ1_S.gguf) | i1-IQ1_S | 7.7 | for the desperate | | [GGUF](https://huggingface.co/mradermacher/Umbra-v3-MoE-4x11b-i1-GGUF/resolve/main/Umbra-v3-MoE-4x11b.i1-IQ2_XXS.gguf) | i1-IQ2_XXS | 9.8 | | | [GGUF](https://huggingface.co/mradermacher/Umbra-v3-MoE-4x11b-i1-GGUF/resolve/main/Umbra-v3-MoE-4x11b.i1-IQ2_XS.gguf) | i1-IQ2_XS | 10.9 | | | [GGUF](https://huggingface.co/mradermacher/Umbra-v3-MoE-4x11b-i1-GGUF/resolve/main/Umbra-v3-MoE-4x11b.i1-IQ2_S.gguf) | i1-IQ2_S | 11.1 | | | [GGUF](https://huggingface.co/mradermacher/Umbra-v3-MoE-4x11b-i1-GGUF/resolve/main/Umbra-v3-MoE-4x11b.i1-IQ2_M.gguf) | i1-IQ2_M | 12.2 | | | [GGUF](https://huggingface.co/mradermacher/Umbra-v3-MoE-4x11b-i1-GGUF/resolve/main/Umbra-v3-MoE-4x11b.i1-Q2_K.gguf) | i1-Q2_K | 13.4 | IQ3_XXS probably better | | [GGUF](https://huggingface.co/mradermacher/Umbra-v3-MoE-4x11b-i1-GGUF/resolve/main/Umbra-v3-MoE-4x11b.i1-IQ3_XXS.gguf) | i1-IQ3_XXS | 14.2 | lower quality | | [GGUF](https://huggingface.co/mradermacher/Umbra-v3-MoE-4x11b-i1-GGUF/resolve/main/Umbra-v3-MoE-4x11b.i1-IQ3_XS.gguf) | i1-IQ3_XS | 14.9 | | | [GGUF](https://huggingface.co/mradermacher/Umbra-v3-MoE-4x11b-i1-GGUF/resolve/main/Umbra-v3-MoE-4x11b.i1-Q3_K_S.gguf) | i1-Q3_K_S | 15.8 | IQ3_XS probably better | | [GGUF](https://huggingface.co/mradermacher/Umbra-v3-MoE-4x11b-i1-GGUF/resolve/main/Umbra-v3-MoE-4x11b.i1-IQ3_S.gguf) | i1-IQ3_S | 15.8 | beats Q3_K* | | [GGUF](https://huggingface.co/mradermacher/Umbra-v3-MoE-4x11b-i1-GGUF/resolve/main/Umbra-v3-MoE-4x11b.i1-IQ3_M.gguf) | i1-IQ3_M | 16.1 | | | [GGUF](https://huggingface.co/mradermacher/Umbra-v3-MoE-4x11b-i1-GGUF/resolve/main/Umbra-v3-MoE-4x11b.i1-Q3_K_M.gguf) | i1-Q3_K_M | 17.5 | IQ3_S probably better | | [GGUF](https://huggingface.co/mradermacher/Umbra-v3-MoE-4x11b-i1-GGUF/resolve/main/Umbra-v3-MoE-4x11b.i1-Q3_K_L.gguf) | i1-Q3_K_L | 19.0 | IQ3_M probably better | | [GGUF](https://huggingface.co/mradermacher/Umbra-v3-MoE-4x11b-i1-GGUF/resolve/main/Umbra-v3-MoE-4x11b.i1-Q4_K_S.gguf) | i1-Q4_K_S | 20.8 | optimal size/speed/quality | | [GGUF](https://huggingface.co/mradermacher/Umbra-v3-MoE-4x11b-i1-GGUF/resolve/main/Umbra-v3-MoE-4x11b.i1-Q4_K_M.gguf) | i1-Q4_K_M | 22.1 | fast, recommended | | [GGUF](https://huggingface.co/mradermacher/Umbra-v3-MoE-4x11b-i1-GGUF/resolve/main/Umbra-v3-MoE-4x11b.i1-Q5_K_S.gguf) | i1-Q5_K_S | 25.1 | | | [GGUF](https://huggingface.co/mradermacher/Umbra-v3-MoE-4x11b-i1-GGUF/resolve/main/Umbra-v3-MoE-4x11b.i1-Q5_K_M.gguf) | i1-Q5_K_M | 25.9 | | Here is a handy graph by ikawrakow comparing some lower-quality quant types (lower is better): ![image.png](https://www.nethype.de/huggingface_embed/quantpplgraph.png) And here are Artefact2's thoughts on the matter: https://gist.github.com/Artefact2/b5f810600771265fc1e39442288e8ec9 ## Thanks I thank my company, [nethype GmbH](https://www.nethype.de/), for letting me use its servers and providing upgrades to my workstation to enable this work in my free time.