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.

InternScience logo

Can the RTX 5090 run MLExpert?

MLExpert·34.7B (3.5B active)·GeForce RTX 5090·32 GB VRAM

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

Yes. The RTX 5090 runs MLExpert fully in VRAM at Q5_K_M (about 28 GB of 32 GB), at roughly 401 tokens/s.

best quant
Q5_K_M
needs
28 GB
usable
32 GB
est. speed
~401 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 RTX 5090

QuantQualityMemoryFitSpeed
Q2_K Heavy loss15 GBfits~868 tok/s
Q3_K_M Noticeable loss18 GBfits~584 tok/s
Q4_K_M Small loss23 GBfits~471 tok/s
Q5_K_M pickNear-lossless28 GBfits~401 tok/s
Q6_K Near-lossless31 GBtight~346 tok/s
Q8_0 Lossless37 GBoffload~53 tok/s
FP16 / BF16 Lossless69 GBtoo big—

Speed is a bandwidth-based estimate for generating text (1792 GB/s); prompt processing and long contexts are slower.

Measured speed on the RTX 5090

No community reports for this pair yet.

Report your own speed →

Run MLExpert on the RTX 5090

Using Q5_K_M, the best fit for RTX 5090.

Ollama · Q5_K_M straight from Hugging Facedocs
ollama run hf.co/bartowski/InternScience_MLExpert-GGUF:Q5_K_M
llama.cpp · OpenAI-compatible server with a web UIdocs
llama-server -hf bartowski/InternScience_MLExpert-GGUF:Q5_K_M

Downloads the GGUF on first run. Installed through llama.app? The same flags work with `llama serve`.

LM Studio · Download with the lms CLIdocs
lms get bartowski/InternScience_MLExpert-GGUF@Q5_K_M

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

vLLM · Serve an OpenAI-compatible APIdocs
vllm serve InternScience/MLExpert
vLLM · Docker Compose (NVIDIA)docs
services:
  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 an OpenAI-compatible APIdocs
sglang serve InternScience/MLExpert --host 0.0.0.0 --port 30000
SGLang · Docker Compose (NVIDIA)docs
services:
  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.

Transformers · Chat with the pipeline APIdocs
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 RTX 5090 comfortably

All that fit →
ModelParamsBest quantMemorySpeedLicence
36BA3BQ5_K_M≈26 GBFits in memory~462 tok/sCommercial OK
31.2BA4.6BQ5_K_M≈29 GBFits in memory~301 tok/sCommercial OK
32.8BQ6_K≈29 GBFits in memory~37 tok/sCommercial OK
RecursiveSI logoflintRecursiveSI
39.8BA1.3BQ5_K_M≈29 GBFits in memory~1108 tok/sCommercial OK
20.9BA4.2BQ8_0≈13 GBFits in memory~221 tok/sCommercial OK
30.5BA3BQ6_K≈26 GBFits in memory~399 tok/sCommercial OK

Other GPUs that run MLExpert

Frequently asked questions

Can the RTX 5090 run MLExpert?
Yes. The RTX 5090 runs MLExpert fully in VRAM at Q5_K_M (about 28 GB of 32 GB), at roughly 401 tokens/s.
How fast does MLExpert run on the RTX 5090?
An estimate from memory bandwidth gives about 401 tokens/s at Q5_K_M.
Which quantisation of MLExpert should I use on the RTX 5090?
Q5_K_M: 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.