--- exported_from: microsoft/WizardLM-2-8x22B language: - en library_name: transformers quantized_by: mradermacher --- ## About weighted/imatrix quants of https://huggingface.co/microsoft/WizardLM-2-8x22B static quants are available at https://huggingface.co/mradermacher/WizardLM-2-8x22B-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/WizardLM-2-8x22B-i1-GGUF/resolve/main/WizardLM-2-8x22B.i1-IQ2_XXS.gguf) | i1-IQ2_XXS | 38.0 | | | [GGUF](https://huggingface.co/mradermacher/WizardLM-2-8x22B-i1-GGUF/resolve/main/WizardLM-2-8x22B.i1-IQ2_M.gguf) | i1-IQ2_M | 46.8 | | | [PART 1](https://huggingface.co/mradermacher/WizardLM-2-8x22B-i1-GGUF/resolve/main/WizardLM-2-8x22B.i1-Q2_K.gguf.part1of2) [PART 2](https://huggingface.co/mradermacher/WizardLM-2-8x22B-i1-GGUF/resolve/main/WizardLM-2-8x22B.i1-Q2_K.gguf.part2of2) | i1-Q2_K | 52.2 | IQ3_XXS probably better | | [PART 1](https://huggingface.co/mradermacher/WizardLM-2-8x22B-i1-GGUF/resolve/main/WizardLM-2-8x22B.i1-IQ3_XXS.gguf.part1of2) [PART 2](https://huggingface.co/mradermacher/WizardLM-2-8x22B-i1-GGUF/resolve/main/WizardLM-2-8x22B.i1-IQ3_XXS.gguf.part2of2) | i1-IQ3_XXS | 55.0 | lower quality | | [PART 1](https://huggingface.co/mradermacher/WizardLM-2-8x22B-i1-GGUF/resolve/main/WizardLM-2-8x22B.i1-IQ3_XS.gguf.part1of2) [PART 2](https://huggingface.co/mradermacher/WizardLM-2-8x22B-i1-GGUF/resolve/main/WizardLM-2-8x22B.i1-IQ3_XS.gguf.part2of2) | i1-IQ3_XS | 58.3 | | | [PART 1](https://huggingface.co/mradermacher/WizardLM-2-8x22B-i1-GGUF/resolve/main/WizardLM-2-8x22B.i1-IQ3_S.gguf.part1of2) [PART 2](https://huggingface.co/mradermacher/WizardLM-2-8x22B-i1-GGUF/resolve/main/WizardLM-2-8x22B.i1-IQ3_S.gguf.part2of2) | i1-IQ3_S | 61.6 | beats Q3_K* | | [PART 1](https://huggingface.co/mradermacher/WizardLM-2-8x22B-i1-GGUF/resolve/main/WizardLM-2-8x22B.i1-Q3_K_S.gguf.part1of2) [PART 2](https://huggingface.co/mradermacher/WizardLM-2-8x22B-i1-GGUF/resolve/main/WizardLM-2-8x22B.i1-Q3_K_S.gguf.part2of2) | i1-Q3_K_S | 61.6 | IQ3_XS probably better | | [PART 1](https://huggingface.co/mradermacher/WizardLM-2-8x22B-i1-GGUF/resolve/main/WizardLM-2-8x22B.i1-IQ3_M.gguf.part1of2) [PART 2](https://huggingface.co/mradermacher/WizardLM-2-8x22B-i1-GGUF/resolve/main/WizardLM-2-8x22B.i1-IQ3_M.gguf.part2of2) | i1-IQ3_M | 64.6 | | | [PART 1](https://huggingface.co/mradermacher/WizardLM-2-8x22B-i1-GGUF/resolve/main/WizardLM-2-8x22B.i1-Q3_K_M.gguf.part1of2) [PART 2](https://huggingface.co/mradermacher/WizardLM-2-8x22B-i1-GGUF/resolve/main/WizardLM-2-8x22B.i1-Q3_K_M.gguf.part2of2) | i1-Q3_K_M | 67.9 | IQ3_S probably better | | [PART 1](https://huggingface.co/mradermacher/WizardLM-2-8x22B-i1-GGUF/resolve/main/WizardLM-2-8x22B.i1-Q3_K_L.gguf.part1of2) [PART 2](https://huggingface.co/mradermacher/WizardLM-2-8x22B-i1-GGUF/resolve/main/WizardLM-2-8x22B.i1-Q3_K_L.gguf.part2of2) | i1-Q3_K_L | 72.7 | IQ3_M probably better | | [PART 1](https://huggingface.co/mradermacher/WizardLM-2-8x22B-i1-GGUF/resolve/main/WizardLM-2-8x22B.i1-IQ4_XS.gguf.part1of2) [PART 2](https://huggingface.co/mradermacher/WizardLM-2-8x22B-i1-GGUF/resolve/main/WizardLM-2-8x22B.i1-IQ4_XS.gguf.part2of2) | i1-IQ4_XS | 75.6 | | | [PART 1](https://huggingface.co/mradermacher/WizardLM-2-8x22B-i1-GGUF/resolve/main/WizardLM-2-8x22B.i1-Q4_0.gguf.part1of2) [PART 2](https://huggingface.co/mradermacher/WizardLM-2-8x22B-i1-GGUF/resolve/main/WizardLM-2-8x22B.i1-Q4_0.gguf.part2of2) | i1-Q4_0 | 80.0 | fast, low quality | | [PART 1](https://huggingface.co/mradermacher/WizardLM-2-8x22B-i1-GGUF/resolve/main/WizardLM-2-8x22B.i1-Q4_K_S.gguf.part1of2) [PART 2](https://huggingface.co/mradermacher/WizardLM-2-8x22B-i1-GGUF/resolve/main/WizardLM-2-8x22B.i1-Q4_K_S.gguf.part2of2) | i1-Q4_K_S | 80.6 | optimal size/speed/quality | | [PART 1](https://huggingface.co/mradermacher/WizardLM-2-8x22B-i1-GGUF/resolve/main/WizardLM-2-8x22B.i1-Q4_K_M.gguf.part1of2) [PART 2](https://huggingface.co/mradermacher/WizardLM-2-8x22B-i1-GGUF/resolve/main/WizardLM-2-8x22B.i1-Q4_K_M.gguf.part2of2) | i1-Q4_K_M | 85.7 | fast, recommended | | [PART 1](https://huggingface.co/mradermacher/WizardLM-2-8x22B-i1-GGUF/resolve/main/WizardLM-2-8x22B.i1-Q5_K_S.gguf.part1of2) [PART 2](https://huggingface.co/mradermacher/WizardLM-2-8x22B-i1-GGUF/resolve/main/WizardLM-2-8x22B.i1-Q5_K_S.gguf.part2of2) | i1-Q5_K_S | 97.1 | | | [PART 1](https://huggingface.co/mradermacher/WizardLM-2-8x22B-i1-GGUF/resolve/main/WizardLM-2-8x22B.i1-Q5_K_M.gguf.part1of3) [PART 2](https://huggingface.co/mradermacher/WizardLM-2-8x22B-i1-GGUF/resolve/main/WizardLM-2-8x22B.i1-Q5_K_M.gguf.part2of3) [PART 3](https://huggingface.co/mradermacher/WizardLM-2-8x22B-i1-GGUF/resolve/main/WizardLM-2-8x22B.i1-Q5_K_M.gguf.part3of3) | i1-Q5_K_M | 100.1 | | | [PART 1](https://huggingface.co/mradermacher/WizardLM-2-8x22B-i1-GGUF/resolve/main/WizardLM-2-8x22B.i1-Q6_K.gguf.part1of3) [PART 2](https://huggingface.co/mradermacher/WizardLM-2-8x22B-i1-GGUF/resolve/main/WizardLM-2-8x22B.i1-Q6_K.gguf.part2of3) [PART 3](https://huggingface.co/mradermacher/WizardLM-2-8x22B-i1-GGUF/resolve/main/WizardLM-2-8x22B.i1-Q6_K.gguf.part3of3) | i1-Q6_K | 115.6 | practically like static Q6_K | 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.