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.

Alternatives to MLExpert

120 open-weight models that do the same job as MLExpert (34.7B chat), ranked by how close they are in task and size, then by popularity and how recent they are. Licence and VRAM are shown for each.

the original
InternScience logo

MLExpert

InternScience·Chat

≈22 GB

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.

34.7BA3.5B256K ctx Apache 2.0
3.4KOct 2026gguf
--vram
any
--gpus
--quant
  1. zai-org logo

    GLM-4-32B-0414

    zai-org·Chat

    ≈20 GB

    The GLM family welcomes new members, the GLM-4-32B-0414 series models, featuring 32 billion parameters. Its performance is comparable to OpenAI’s GPT series and DeepSeek’s V3/R1 series. It also supports very user-friendly local deployment features.

    32.6B32K ctx MIT
    8.4KApr 2025gguf
    • Same category
    • Similar size (32.6B vs 34.7B)
    • Commercial use OK
    vs
  2. ByteDance-Seed logo

    Seed-OSS-36B-Base

    ByteDance-Seed·Chat

    ≈24 GB

    Seed-OSS is a series of open-source large language models developed by ByteDance's Seed Team, designed for powerful long-context, reasoning, agent and general capabilities, and versatile developer-friendly features.

    36.2B512K ctx Apache 2.0
    2.4KAug 2025gguf
    • Same category
    • Similar size (36.2B vs 34.7B)
    • Commercial use OK
    vs
  3. nvidia logo≈19 GB

    We're excited to introduce Nemotron-Labs-Audex-30B-A3B, a unified audio-text LLM built on Nemotron-Cascade-2-30B-A3B, a strong text-only MoE LLM with 30B MoE model with 3B activated parameters.

    30BA3Bnvidia-oneway-noncommercial-license
    517Jul 2026gguf
    • Same category
    • Similar size (30B vs 34.7B)
    vs
  4. t-tech logo

    T-pro-it-2.1

    t-tech·Chat

    ≈22 GB

    We introduce the updated version of the T-pro-it-2.0, named T-pro-it-2.1, featuring the following key enhancements:

    32.8B40K ctx Apache 2.0
    1.1KDec 2025gguf
    • Same category
    • Similar size (32.8B vs 34.7B)
    • Commercial use OK
    vs
  5. allenai logo

    FlexOlmo-7x7B-1T

    allenai·Chat

    ≈22 GB

    FlexOlmo is a new kind of LM that unlocks a new paradigm of data collaboration. With FlexOlmo, data owners can contribute to the development of open language models without giving up control of their data.

    33.3B4K ctx Apache 2.0
    3.5KJun 2025
    • Same category
    • Similar size (33.3B vs 34.7B)
    • Commercial use OK
    vs
  6. allenai logo≈21 GB

    OLMo 2 32B Instruct March 2025 is post-trained variant of the OLMo-2 32B March 2025 model, which has undergone supervised finetuning on an OLMo-specific variant of the Tülu 3 dataset, further DPO training on this dataset, and final RLVR training on this dataset.

    32.2B4K ctx Apache 2.0
    9.2KMar 2025gguf
    • Same category
    • Similar size (32.2B vs 34.7B)
    • Commercial use OK
    vs
  7. ibm-granite logo

    granite-4.1-30b-base

    ibm-granite·Chat

    ≈20 GB

    Model Summary: Granite‑4.1‑30B‑Base is a decoder‑only language model with long‑context capabilities, designed to support a broad range of text‑to‑text generation tasks.

    28.9B128K ctx Apache 2.0
    1.2KApr 2026gguf
    • Same category
    • Similar size (28.9B vs 34.7B)
    • Commercial use OK
    vs
  8. tiiuae logo≈22 GB

    For more details about the training protocol of this model, please refer to the Falcon-H1 technical blogpost and Technical Report.

    33.6B256K ctx Falcon LLM License
    3.9KMay 2025gguf
    • Same category
    • Similar size (33.6B vs 34.7B)
    vs
  9. Skywork logo

    MindLink-32B-0801

    Skywork·Chat

    ≈22 GB

    We introduce MindLink, a new family of large language models developed by Kunlun Inc. Built on Qwen, these models incorporate our latest advances in post-training techniques. MindLink demonstrates strong performance across various common benchmarks and is widely applicable in diverse AI scenarios.

    32.8B40K ctx Apache 2.0
    1.8KAug 2025gguf
    • Same category
    • Similar size (32.8B vs 34.7B)
    • Commercial use OK
    vs
  10. FINAL-Bench logo

    Darwin-27B-RSI

    FINAL-Bench·Chat

    ≈18 GB

    Qwen3.5-27B family · 27B dense · Thinking mode · BF16 · Apache 2.0 No human-written answers. The model generated its own learning signal — and got measurably better.

    26.9B256K ctx Apache 2.0
    388Sept 2026gguf
    • Same category
    • Similar size (26.9B vs 34.7B)
    • Commercial use OK
    vs
  11. google logo≈20 GB

    This section describes the MedGemma model and how to use it.

    27B128K ctx health-ai-developer-foundations
    14.9KMay 2025gguf
    • Same category
    • Similar size (27B vs 34.7B)
    vs
  12. allenai logo

    OLMo-2-0325-32B

    allenai·Chat

    ≈21 GB

    We introduce OLMo 2 32B, the largest model in the OLMo 2 family. OLMo 2 was pre-trained on OLMo-mix-1124 and uses Dolmino-mix-1124 for mid-training.

    32.2B4K ctx Apache 2.0
    5KFeb 2025gguf
    • Same category
    • Similar size (32.2B vs 34.7B)
    • Commercial use OK
    vs
  13. Qwen logo

    Qwen2.5-32B

    Qwen·Chat

    ≈22 GB

    Qwen2.5 is the latest series of Qwen large language models. For Qwen2.5, we release a number of base language models and instruction-tuned language models ranging from 0.5 to 72 billion parameters. Qwen2.5 brings the following improvements upon Qwen2:

    32.8B128K ctx Apache 2.0
    13.9KSept 2024gguf
    • Same category
    • Similar size (32.8B vs 34.7B)
    • Commercial use OK
    vs
  14. zai-org logo≈22 GB

    The GLM family welcomes a new generation of open-source models, the GLM-4-32B-0414 series, featuring 32 billion parameters. Its performance is comparable to OpenAI's GPT series and DeepSeek's V3/R1 series, and it supports very user-friendly local deployment features.

    33.1B128K ctx MIT
    2.5KApr 2025gguf
    • Same category
    • Similar size (33.1B vs 34.7B)
    • Commercial use OK
    vs
  15. inclusionAI logo

    Rubicon-Preview

    inclusionAI·Chat

    ≈19 GB

    This is the model card for Rubicon-preview, a 30B-A3B parameter model trained with a novel reinforcement learning framework using "rubric anchors" to excel at open-ended, creative, and humanities-centric tasks.

    30.5BA3.4B40K ctx Apache 2.0
    1.9KAug 2025gguf
    • Same category
    • Similar size (30.5B vs 34.7B)
    • Commercial use OK
    vs
  16. LiquidAI logo

    LFM2-24B-A2B

    LiquidAI·Chat

    ≈15 GB

    LFM2 is a family of hybrid models designed for on-device deployment. LFM2-24B-A2B is the largest model in the family, scaling the architecture to 24 billion parameters while keeping inference efficient.

    23.8BA2B125K ctx lfm1.0
    4KFeb 2026ollama
    • Same category
    • Similar size (23.8B vs 34.7B)
    vs
  17. openai logo≈14 GB

    gpt-oss-safeguard-120b and gpt-oss-safeguard-20b are safety reasoning models built-upon gpt-oss. With these models, you can classify text content based on safety policies that you provide and perform a suite of foundational safety tasks. These models are intended for safety use cases.

    21.5BA4.8B128K ctx Apache 2.0
    27.2KSept 2025ollama
    • Same category
    • Similar size (21.5B vs 34.7B)
    • Commercial use OK
    vs
  18. 01-ai logo

    Yi-1.5-34B-Chat

    01-ai·Chat

    ≈22 GB

    Yi-1.5 is an upgraded version of Yi. It is continuously pre-trained on Yi with a high-quality corpus of 500B tokens and fine-tuned on 3M diverse fine-tuning samples.

    34.4B4K ctx Apache 2.0
    15.1KMay 2024gguf
    • Same category
    • Similar size (34.4B vs 34.7B)
    • Commercial use OK
    vs
  19. ByteDance-Seed logo

    Seed-OSS-36B-Base-woSyn

    ByteDance-Seed·Chat

    ≈24 GB

    Seed-OSS is a series of open-source large language models developed by ByteDance's Seed Team, designed for powerful long-context, reasoning, agent and general capabilities, and versatile developer-friendly features.

    36.2B512K ctx Apache 2.0
    551Aug 2025gguf
    • Same category
    • Similar size (36.2B vs 34.7B)
    • Commercial use OK
    vs
  20. Llama-3.3-Nemotron-Super-49B-v1 is a large language model (LLM) which is a derivative of Meta Llama-3.3-70B-Instruct (AKA the reference model). It is a reasoning model that is post trained for reasoning, human chat preferences, and tasks, such as RAG and tool calling.

    49.9B128K ctx NVIDIA Open Model License
    35.9KMar 2025gguf
    • Same category
    • Similar size (49.9B vs 34.7B)
    • Commercial use OK
    vs
  21. t-tech logo

    T-pro-it-2.0

    t-tech·Chat

    ≈22 GB

    T-pro-it-2.0 is a model built upon the Qwen 3 model family and incorporates both continual pre-training and alignment techniques.

    32.8B40K ctx Apache 2.0
    843Jul 2025gguf
    • Same category
    • Similar size (32.8B vs 34.7B)
    • Commercial use OK
    vs
  22. utter-project logo

    EuroLLM-22B-Instruct-2512

    utter-project·Chat

    ≈16 GB

    This is the model card for EuroLLM-22B-Instruct. You can also check the pre-trained version: EuroLLM-22B-2515.

    22.6B32K ctx Apache 2.0
    6.7KDec 2025gguf
    • Same category
    • Similar size (22.6B vs 34.7B)
    • Commercial use OK
    vs
  23. NousResearch logo

    nomos-1

    NousResearch·Chat

    ≈19 GB

    We release Nomos 1, a specialization of Qwen/Qwen3-30B-A3B-Thinking-2507 for mathematical problem-solving and proof-writing in natural language. Nomos-1 was trained in collaboration with Hillclimb AI.

    30.5BA3.4B256K ctx Apache 2.0
    409Dec 2025gguf
    • Same category
    • Similar size (30.5B vs 34.7B)
    • Commercial use OK
    vs
  24. Llama-3.3-Nemotron-Super-49B-v1.5 is a significantly upgraded version of Llama-3.3-Nemotron-Super-49B-v1 and is a large language model (LLM) which is a derivative of Meta Llama-3.3-70B-Instruct (AKA the reference model).

    49.9B128K ctx NVIDIA Open Model License
    9.9KJul 2025gguf
    • Same category
    • Similar size (49.9B vs 34.7B)
    • Commercial use OK
    vs