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

MLExpert

InternScience·Instruct·released 1 Oct 2026

Apache 2.034.7BA3.5BMoE256K ctx ≈22 GB
Your GPU
ollama run hf.co/bartowski/InternScience_MLExpert-GGUF:Q4_K_M
All 6 runners ↓

MLExpert is a 35B Mixture-of-Experts agentic model for autonomous machine learning engineering. It organizes experiments, writes and revises code, and improves solutions through sustained interaction with execution environments.

Notify me when

How much VRAM does MLExpert need?

At Q4_K_M MLExpert needs about 22 GB with an 8K context (22 GB at Q4, 69 GB at full precision). Pick your GPU to see what fits and how fast it runs.

--vram
any
--gpus
--quant

Needs: ≈23 GB · Q4_K_M · 8K context — Pick a GPU to check the fit.

Change quant, context or KV cache to compare
Advanced: context 8K · KV cache 16-bit
Context
KV cache

Smallest GPU for each quant and context (KV cache 16-bit)

Quantisation8K16K32K64K128K256K
Q3_K_M
Q4_K_M
Q5_K_M
Q6_K
Q8_0
FP16

≤16 GB card24–32 GB48–96 GBmore

QuantisationQualityWeightsKV cacheTotal
Q2_KHeavy lossSmallest, noticeably worse qualityHeavy loss13 GB file0.6 GB≈15 GB
Q3_K_MNoticeable lossSmall, some quality lossNoticeable loss16 GB file0.6 GB≈18 GB
Q4_K_MrecommendedSmall lossRecommended default for local useSmall loss21 GB file0.6 GB≈23 GB
Q5_K_MNear-losslessNear-lossless for most tasksNear-lossless25 GB file0.6 GB≈28 GB
Q6_KNear-losslessPractically losslessNear-lossless29 GB file0.6 GB≈31 GB
Q8_0LosslessLossless in practice, twice the size of Q4Lossless34 GB file0.6 GB≈37 GB
FP16 / BF16LosslessFull precision, as releasedLossless65 GB file0.6 GB≈69 GB

Estimates for 8K tokens of context: weights + KV cache + runtime overhead. All experts stay in memory, but only the active ones are read per token. Speed assumes memory-bound decoding at your GPU’s bandwidth. Pick a GPU above and the fit column checks every row against it. Weights marked “file” use the real GGUF download size; “est.” rows are computed from the parameter count.

How to pick a quant

Take the biggest quant that fits with room left for context. Q4_K_M is the usual sweet spot. Below Q4 quality drops fast, above Q6 you mostly pay memory for nothing. Big models (70B+, large MoE) tolerate low quants better than small ones.

  • LosslessF16, Q8_0
  • Near-losslessQ6_K, Q5_K_M
  • Small lossQ5_K_S, Q4_K_M, IQ4_NL, IQ4_XS, Q4_K_S
  • Noticeable lossQ4_0, Q3_K_L, Q3_K_M, IQ3_M
  • Heavy lossIQ3_XXS, Q3_K_S, IQ2_M, Q2_K

Typical quality loss, not measured per model. Tiers come from KL divergence against FP16 on Llama 3 8B: llama.cpp perplexity README: LLaMA 3 8B scoreboard (KLD and ΔPPL vs FP16, Wikitext-2); llama.cpp quantize tool: ΔPPL per quant type @ Llama-3-8B

How much context fits?

Model quant
KV cache

At 8K context, Q4_K_M needs ≈23 GB(21 GB weights + 0.6 GB KV cache + 1.5 GB overhead). Pick a GPU above to see how much context fits.

010203040501K2K4K8K16K32K64K128K256KGB

Weights + overheadKV cache

Can the Instinct MI50 32GB run MLExpert? →

Speed vs quality per quant

Smaller quants run faster and fit in less memory, but lose quality.

Q2_KQ3_K_MQ4_K_MQ5_K_MQ6_KQ8_0F16

Fits in VRAMTight fitCPU offload✕ won't runQuality is typical KL divergence vs FP16, not measured for this model.

Real-world speed of MLExpert

Tokens per second measured by people running it at home. A GPU shows a median once 3 reports agree on the same setup. Reports far from our estimate wait for a review. Measure yours with llama-bench →

No medians yet. Be one of the first 3 to report a setup.

All reports

Run MLExpert locally

--quantQ4_K_Mdefault; pick your GPU to tailor itbartowski/InternScience_MLExpert-GGUF

Q4_K_M straight from Hugging Facedocs
ollama run hf.co/bartowski/InternScience_MLExpert-GGUF:Q4_K_M

Install Ollama

OpenAI-compatible server with a web UIdocs
llama-server -hf bartowski/InternScience_MLExpert-GGUF:Q4_K_M

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

Install llama.cpp

Download with the lms CLIdocs
lms get bartowski/InternScience_MLExpert-GGUF@Q4_K_M

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

Install LM Studio

Serve an OpenAI-compatible APIdocs
vllm serve InternScience/MLExpert
Docker Compose (NVIDIA)docs
yaml
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.

Install vLLM

Serve an OpenAI-compatible APIdocs
sglang serve InternScience/MLExpert --host 0.0.0.0 --port 30000
Docker Compose (NVIDIA)docs
yaml
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.

Install SGLang

Chat with the pipeline APIdocs
python
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"])

Install Transformers

Quantised versions and downloads

About MLExpert

MLExpert is a 35B Mixture-of-Experts agentic model for autonomous machine learning engineering. It organizes experiments, writes and revises code, and improves solutions through sustained interaction with execution environments. Hierarchical agentic SFT develops general tool-use and domain-specific optimization skills, while Windowed Interaction Reinforcement Learning (WIRL) trains multi-turn decisions using feedback from completed experiments. • Long-horizon optimization: Plans experiments, interprets execution feedback, revisits previous solutions, and selects submission candidates within a persistent workspace. • Strong MLE performance: Achieves a 72.7% medal rate on MLE-Bench Lite, matching GLM-5.3 and Kimi-K3 under the MLExpert Harness. • General-purpose harness compatibility: Achieves a 69.7% medal rate with Claude Code, compared with 42.4% for its base model. • Scientific task optimization: Matches published scientific results on seven of ten NatureBench Lite tasks and surpasses them on five.

Full model card on Hugging Face →

Alternatives to MLExpert

All alternatives →
ornith-ai logo

Ornith-1.0-35B

ornith-ai·Chat

≈22 GB

Aloha! 🌺 Today, we are releasing Ornith-1.0, a self-improving family of open-source models for agentic coding.

35BA3.8B256K ctx MIT
1.5MJun 2026gguf

NVIDIA Nemotron™ is a family of open models with open weights, training data, and recipes, delivering leading efficiency and accuracy for building specialized AI agents.

31.6BA3B256K ctx openmdw-1.1
963.1KAug 2026gguf
farbodtavakkoli logo

OTel-2.0-LLM-31B-IT

farbodtavakkoli·Chat

≈27 GB

Versioning notice: This card describes the current OTel 2.0 OSFT checkpoint. The model weights may be updated over time. For reproducible evaluation or production deployment, pin a specific model revision, checkpoint hash, or release tag rather than relying on the floating latest version.

31.3B256K ctx Apache 2.0
866.9KJul 2026gguf
poolside logo

Laguna-XS-2.1

poolside·Chat

≈22 GB

Laguna XS 2.1 is a 33B total parameter Mixture-of-Experts model with 3B activated parameters per token designed for agentic coding and long-horizon work on a local machine.

33.4BA2.2B256K ctx Custom licence
587.1KJun 2026ollama
ornith-ai logo

Ornith-1.5-35B-A3B

ornith-ai·Chat

≈22 GB

Chirp Chirp! 🐦 We are introducing Ornith-1.5, a major step toward building foundation models through end-to-end self-improvement.

36BA3B256K ctx MIT
318.5KAug 2026ollama
poolside logo

Laguna-XS.2

poolside·Chat

≈22 GB

Laguna XS.2 is a 33B total parameter Mixture-of-Experts model with 3B activated parameters per token designed for agentic coding and long-horizon work on a local machine. It uses Sliding Window Attention with per-head gating in 30 out of 40 layers for fast inference and low KV cache requirements.

33.4BA2.2B256K ctx Apache 2.0
731KApr 2026ollama

MLExpert FAQ

How much VRAM does MLExpert need?
About 22 GB at Q4_K_M with an 8K context. Q8_0 needs about 37 GB and full precision about 69 GB. Longer contexts need more memory for the KV cache.
Can MLExpert run on the RTX 5090?
Yes. It fits at Q5_K_M (about 28 GB), at an estimated 401 tokens/s. It fits comfortably on the Instinct MI50 32GB, Apple M4 Pro (48 GB), RTX A6000.
Can I use MLExpert commercially?
Yes. It is released under Apache 2.0. Permissive: use, modify and sell freely; keep the notice and licence text.
What is the context length of MLExpert?
262,144 tokens (256K). The full window needs far more memory than an 8K context.
How do I run MLExpert with Ollama or llama.cpp?
With Ollama: ollama run hf.co/bartowski/InternScience_MLExpert-GGUF:Q4_K_M. With llama.cpp: llama-server -hf bartowski/InternScience_MLExpert-GGUF:Q4_K_M, which starts an OpenAI-compatible server.

Data from Hugging Face, the model card and public leaderboards. Memory and speed figures are estimates; real numbers depend on the runtime, context length and drivers. Always check the licence yourself before commercial use.