--- language: - en library_name: transformers license: apache-2.0 quantized_by: mradermacher --- ## About weighted/imatrix quants of https://huggingface.co/mistralai/Mixtral-8x7B-Instruct-v0.1 Artefact2 also made some, with a different weight matrix, at https://huggingface.co/Artefact2/Mixtral-8x7B-Instruct-v0.1-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/Mixtral-8x7B-Instruct-v0.1-i1-GGUF/resolve/main/Mixtral-8x7B-Instruct-v0.1.i1-IQ1_S.gguf) | i1-IQ1_S | 9.8 | for the desperate | | [GGUF](https://huggingface.co/mradermacher/Mixtral-8x7B-Instruct-v0.1-i1-GGUF/resolve/main/Mixtral-8x7B-Instruct-v0.1.i1-IQ2_XXS.gguf) | i1-IQ2_XXS | 12.6 | | | [GGUF](https://huggingface.co/mradermacher/Mixtral-8x7B-Instruct-v0.1-i1-GGUF/resolve/main/Mixtral-8x7B-Instruct-v0.1.i1-IQ2_XS.gguf) | i1-IQ2_XS | 13.9 | | | [GGUF](https://huggingface.co/mradermacher/Mixtral-8x7B-Instruct-v0.1-i1-GGUF/resolve/main/Mixtral-8x7B-Instruct-v0.1.i1-IQ2_S.gguf) | i1-IQ2_S | 14.4 | | | [GGUF](https://huggingface.co/mradermacher/Mixtral-8x7B-Instruct-v0.1-i1-GGUF/resolve/main/Mixtral-8x7B-Instruct-v0.1.i1-IQ2_M.gguf) | i1-IQ2_M | 15.8 | | | [GGUF](https://huggingface.co/mradermacher/Mixtral-8x7B-Instruct-v0.1-i1-GGUF/resolve/main/Mixtral-8x7B-Instruct-v0.1.i1-Q2_K.gguf) | i1-Q2_K | 17.6 | IQ3_XXS probably better | | [GGUF](https://huggingface.co/mradermacher/Mixtral-8x7B-Instruct-v0.1-i1-GGUF/resolve/main/Mixtral-8x7B-Instruct-v0.1.i1-IQ3_XXS.gguf) | i1-IQ3_XXS | 18.6 | fast, lower quality | | [GGUF](https://huggingface.co/mradermacher/Mixtral-8x7B-Instruct-v0.1-i1-GGUF/resolve/main/Mixtral-8x7B-Instruct-v0.1.i1-Q3_K_XS.gguf) | i1-Q3_K_XS | 19.3 | | | [GGUF](https://huggingface.co/mradermacher/Mixtral-8x7B-Instruct-v0.1-i1-GGUF/resolve/main/Mixtral-8x7B-Instruct-v0.1.i1-IQ3_XS.gguf) | i1-IQ3_XS | 19.5 | | | [GGUF](https://huggingface.co/mradermacher/Mixtral-8x7B-Instruct-v0.1-i1-GGUF/resolve/main/Mixtral-8x7B-Instruct-v0.1.i1-Q3_K_S.gguf) | i1-Q3_K_S | 20.7 | IQ3_XS probably better | | [GGUF](https://huggingface.co/mradermacher/Mixtral-8x7B-Instruct-v0.1-i1-GGUF/resolve/main/Mixtral-8x7B-Instruct-v0.1.i1-IQ3_S.gguf) | i1-IQ3_S | 20.7 | fast, beats Q3_K* | | [GGUF](https://huggingface.co/mradermacher/Mixtral-8x7B-Instruct-v0.1-i1-GGUF/resolve/main/Mixtral-8x7B-Instruct-v0.1.i1-IQ3_M.gguf) | i1-IQ3_M | 21.7 | | | [GGUF](https://huggingface.co/mradermacher/Mixtral-8x7B-Instruct-v0.1-i1-GGUF/resolve/main/Mixtral-8x7B-Instruct-v0.1.i1-Q3_K_M.gguf) | i1-Q3_K_M | 22.8 | IQ3_S probably better | | [GGUF](https://huggingface.co/mradermacher/Mixtral-8x7B-Instruct-v0.1-i1-GGUF/resolve/main/Mixtral-8x7B-Instruct-v0.1.i1-Q3_K_L.gguf) | i1-Q3_K_L | 24.4 | IQ3_M probably better | | [GGUF](https://huggingface.co/mradermacher/Mixtral-8x7B-Instruct-v0.1-i1-GGUF/resolve/main/Mixtral-8x7B-Instruct-v0.1.i1-Q4_K_S.gguf) | i1-Q4_K_S | 27.0 | optimal size/speed/quality | | [GGUF](https://huggingface.co/mradermacher/Mixtral-8x7B-Instruct-v0.1-i1-GGUF/resolve/main/Mixtral-8x7B-Instruct-v0.1.i1-Q4_K_M.gguf) | i1-Q4_K_M | 28.7 | fast, medium quality | | [GGUF](https://huggingface.co/mradermacher/Mixtral-8x7B-Instruct-v0.1-i1-GGUF/resolve/main/Mixtral-8x7B-Instruct-v0.1.i1-Q5_K_S.gguf) | i1-Q5_K_S | 32.5 | | | [GGUF](https://huggingface.co/mradermacher/Mixtral-8x7B-Instruct-v0.1-i1-GGUF/resolve/main/Mixtral-8x7B-Instruct-v0.1.i1-Q5_K_M.gguf) | i1-Q5_K_M | 33.5 | | 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.