MosaicML
MPT-30B-Instruct (30B parameters) requires approximately 46.8 GB of VRAM with Q5_K_M quantization. For the best balance of quality and speed, we recommend hardware with at least 54 GB of VRAM.
Get started
— copy & paste to run locallyCopy-paste commands to run MPT-30B-Instruct on your machine.
Run
docker run --rm -it ghcr.io/ggerganov/llama.cpp:full \
--hf-repo "mosaicml/mpt-30b-instruct" \
--hf-file "mpt-30b-instruct-Q5_K_M.gguf" \
-c 4096 -ngl 99Quick specs
About this model
Related models
Quick picks
Best hardware
Run this model
Quantization options
No hardware detected — fit column shows raw VRAM estimates
| Quant | Bits | VRAM | Quality | Fit |
|---|---|---|---|---|
Q2_K | 2 | 11.7 GB | Low | — |
Q3_K_S | 3 | 14.7 GB | Low | — |
NVFP4 | 4 | 16.8 GB | Medium | — |
Q4_K_M | 4 | 18.3 GB | Medium | — |
Q5_K_M | 5 | 21.6 GB | High | — |
Q6_K | 6 | 24.6 GB | High | — |
Q8_0 | 8 | 32.1 GB | Very High | — |
F16 | 16 | 61.5 GB | Maximum | — |
Hardware compatibility
Computing compatibility...
Memory breakdown
Frequently asked questions
MPT-30B-Instruct (30B parameters) requires approximately 46.8 GB of VRAM with Q5_K_M quantization. Lower quantizations like Q4_K_M use less memory but may reduce quality.
Yes, MacBook Pro M4 Max 96GB can run MPT-30B-Instruct with a compatibility score of 74/100. It provides 96 GB of memory and achieves approximately 28.4 tokens per second.
The recommended quantization for MPT-30B-Instruct is Q5_K_M, which offers the best balance between model quality and memory efficiency. Higher quantizations preserve more quality but require more VRAM.
The top recommended hardware for MPT-30B-Instruct: NVIDIA H100 80GB (score: 77/100), NVIDIA H800 80GB (score: 77/100), NVIDIA A100 80GB (score: 77/100). These provide the best combination of memory, bandwidth, and compute for running this model locally.
Yes, MPT-30B-Instruct is well-suited for chat as well as reasoning. It was designed with these use cases in mind.
See also
