Can the NVIDIA A100 80GB run MLExpert?
MLExpert·34.7B (3.5B active)·NVIDIA A100 80GB·80 GB VRAM
Yes. The NVIDIA A100 80GB runs MLExpert fully in VRAM at FP16 / BF16 (about 69 GB of 80 GB), at roughly 162 tokens/s.
- best quant
- FP16 / BF16
- needs
- 69 GB
- usable
- 80 GB
- est. speed
- ~162 tok/s
Estimates for an 8K context; offload assumes 32 GB of system RAM.Try other contexts on the MLExpert page →
Every quantisation of MLExpert on the NVIDIA A100 80GB
| Quant | Memory | Fit | Speed |
|---|---|---|---|
| Q2_K | 15 GB | fits | ~987 tok/s |
| Q3_K_M | 18 GB | fits | ~664 tok/s |
| Q4_K_M | 23 GB | fits | ~535 tok/s |
| Q5_K_M | 28 GB | fits | ~456 tok/s |
| Q6_K | 31 GB | fits | ~394 tok/s |
| Q8_0 | 37 GB | fits | ~305 tok/s |
| FP16 / BF16 pick | 69 GB | fits | ~162 tok/s |
Speed is a bandwidth-based estimate for generating text (2039 GB/s); prompt processing and long contexts are slower.
Measured speed on the NVIDIA A100 80GB
No community reports for this pair yet.
Report your own speed →Run MLExpert on the NVIDIA A100 80GB
Using FP16 / BF16, the best fit for NVIDIA A100 80GB.
ollama run hf.co/bartowski/InternScience_MLExpert-GGUF:BF16 llama-server -hf bartowski/InternScience_MLExpert-GGUF:BF16Downloads the GGUF on first run. Installed through llama.app? The same flags work with `llama serve`.
lms get bartowski/InternScience_MLExpert-GGUF@BF16Then load it from the LM Studio app or with `lms load`.
vllm serve InternScience/MLExpertservices:
vllm:
image: vllm/vllm-openai:latest
command: InternScience/MLExpert
ports:
- "8000:8000"
ipc: host
volumes:
- ~/.cache/huggingface:/root/.cache/huggingface
deploy:
resources:
reservations:
devices:
- driver: nvidia
count: all
capabilities: [gpu]API on port 8000.
sglang serve InternScience/MLExpert --host 0.0.0.0 --port 30000services:
sglang:
image: lmsysorg/sglang:latest
command: sglang serve InternScience/MLExpert --host 0.0.0.0 --port 30000
ports:
- "30000:30000"
ipc: host
volumes:
- ~/.cache/huggingface:/root/.cache/huggingface
deploy:
resources:
reservations:
devices:
- driver: nvidia
count: all
capabilities: [gpu]API on port 30000.
from transformers import pipeline
pipe = pipeline("text-generation", model="InternScience/MLExpert", dtype="auto", device_map="auto")
messages = [{"role": "user", "content": "Explain self-hosting in one paragraph."}]
out = pipe(messages, max_new_tokens=512)
print(out[0]["generated_text"][-1]["content"])Chat models that fit the NVIDIA A100 80GB comfortably
All that fit →| Model | Params | Best quant | Memory | Speed |
|---|---|---|---|---|
| 36BA3B | Q8_0 | ≈38 GBFits in memory | ~352 tok/s | |
| 31.2BA4.6B | Q8_0 | ≈39 GBFits in memory | ~229 tok/s | |
| 32.8B | Q8_0 | ≈37 GBFits in memory | ~32 tok/s | |
| 39.8BA1.3B | Q8_0 | ≈43 GBFits in memory | ~844 tok/s | |
| 20.9BA4.2B | Q8_0 | ≈13 GBFits in memory | ~252 tok/s | |
| 30.5BA3B | Q8_0 | ≈33 GBFits in memory | ~352 tok/s |
Other GPUs that run MLExpert
Frequently asked questions
- Can the NVIDIA A100 80GB run MLExpert?
- Yes. The NVIDIA A100 80GB runs MLExpert fully in VRAM at FP16 / BF16 (about 69 GB of 80 GB), at roughly 162 tokens/s.
- How fast does MLExpert run on the NVIDIA A100 80GB?
- An estimate from memory bandwidth gives about 162 tokens/s at FP16 / BF16.
- Which quantisation of MLExpert should I use on the NVIDIA A100 80GB?
- FP16 / BF16: the largest one that fits. Expected quality: 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.