selfhost.directory

The index of self-hostable projects — ranked, categorized and served with live updates.
Own your data and stop giving it to big tech.

NousResearch logo

Can the Instinct MI50 32GB run Arete-27B?

Arete-27B·27.4B·Instinct MI50 32GB·32 GB VRAM

Saved in this browser, used on every model page.
Yes

Yes. The Instinct MI50 32GB runs Arete-27B fully in VRAM at Q6_K (about 26 GB of 32 GB), at roughly 25 tokens/s.

best quant
Q6_K
needs
26 GB
usable
32 GB
est. speed
~25 tok/s

Estimates for an 8K context; offload assumes 32 GB of system RAM.Try other contexts on the Arete-27B page →

Every quantisation of Arete-27B on the Instinct MI50 32GB

QuantQualityMemoryFitSpeed
Q2_K Heavy loss13 GBfits~63 tok/s
Q3_K_M Noticeable loss15 GBfits~42 tok/s
Q4_K_M Small loss19 GBfits~34 tok/s
Q5_K_M Near-lossless23 GBfits~29 tok/s
Q6_K pickNear-lossless26 GBfits~25 tok/s
Q8_0 Lossless31 GBtight~19 tok/s
FP16 / BF16 Lossless55 GBoffload~1.3 tok/s

Speed is a bandwidth-based estimate for generating text (1024 GB/s); prompt processing and long contexts are slower.

Measured speed on the Instinct MI50 32GB

No community reports for this pair yet.

Report your own speed →

Run Arete-27B on the Instinct MI50 32GB

Using Q6_K, the best fit for Instinct MI50 32GB.

Ollama · Q6_K straight from Hugging Facedocs
ollama run hf.co/bartowski/NousResearch_Arete-27B-GGUF:Q6_K

Ollama’s ROCm/CUDA builds don’t cover this GPU. Recent Ollama releases use Vulkan when its Vulkan backend and driver are installed (older ones needed OLLAMA_VULKAN=1); llama.cpp or LM Studio with Vulkan are the safer bet.

llama.cpp · OpenAI-compatible server with a web UIdocs
llama-server -hf bartowski/NousResearch_Arete-27B-GGUF:Q6_K

Needs a Vulkan build of llama.cpp. Downloads the GGUF on first run.

llama.cpp · Docker with Vulkandocs
docker run -d --name llama --device /dev/dri -p 8080:8080 \
  -v ~/.cache/llama.cpp:/root/.cache/llama.cpp \
  ghcr.io/ggml-org/llama.cpp:server-vulkan \
  -hf bartowski/NousResearch_Arete-27B-GGUF:Q6_K --host 0.0.0.0 --port 8080 -ngl 99

Server and web UI on port 8080.

LM Studio · Download with the lms CLIdocs
lms get bartowski/NousResearch_Arete-27B-GGUF@Q6_K

Then load it from the LM Studio app or with `lms load`.

Chat models that fit the Instinct MI50 32GB comfortably

All that fit →
ModelParamsBest quantMemorySpeedLicence
31.2BA4.6BQ5_K_M≈29 GBFits in memory~172 tok/sCommercial OK
20.9BA4.2BQ8_0≈13 GBFits in memory~126 tok/sCommercial OK
32.8BQ6_K≈29 GBFits in memory~21 tok/sCommercial OK
36BA3BQ5_K_M≈26 GBFits in memory~264 tok/sCommercial OK
30.5BA3BQ6_K≈26 GBFits in memory~228 tok/sCommercial OK
24BA2BQ8_0≈26 GBFits in memory~265 tok/sConditions

Other GPUs that run Arete-27B

Frequently asked questions

Can the Instinct MI50 32GB run Arete-27B?
Yes. The Instinct MI50 32GB runs Arete-27B fully in VRAM at Q6_K (about 26 GB of 32 GB), at roughly 25 tokens/s.
How fast does Arete-27B run on the Instinct MI50 32GB?
An estimate from memory bandwidth gives about 25 tokens/s at Q6_K.
Which quantisation of Arete-27B should I use on the Instinct MI50 32GB?
Q6_K: the largest one that fits. Expected quality: near-lossless compared with the full-precision weights.

Memory and speed are estimates from the model's size and the card's memory bandwidth; real numbers depend on the runtime, drivers and context length. Community figures are medians of user reports.