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. badtheorylabs logo

    BTL-4

    badtheorylabs·Chat

    ≈22 GB

    A 35B agentic reasoning model from Bad Theory Labs, fine-tuned from Ornith-1.0-35B on an execution-gated reasoning corpus.

    35.1BA3.9B256K ctx Apache 2.0
    476Aug 2026gguf
    • Same category
    • Similar size (35.1B vs 34.7B)
    • Commercial use OK
    vs
  2. nex-agi logo

    Nex-N2-mini

    nex-agi·Chat

    ≈22 GB

    Today, we are officially releasing and open-sourcing our next-generation model, Nex-N2 — an agent model built for real-world productivity scenarios.

    35.1BA3.9B256K ctx Apache 2.0
    684Jun 2026gguf
    • Same category
    • Similar size (35.1B vs 34.7B)
    • Commercial use OK
    vs
  3. LGAI-EXAONE logo

    EXAONE-4.0-32B

    LGAI-EXAONE·Chat

    ≈21 GB

    🎉 License Updated! We are pleased to announce our more flexible licensing terms 🤗 ✈️ Try on FriendliAI (licensed under commercial purposes) 📢 EXAONE 4.0 is officially supported by HuggingFace transformers! Please check out the guide below

    32B128K ctx exaone
    13.5KJul 2025gguf
    • Same category
    • Similar size (32B vs 34.7B)
    vs
  4. Alibaba-NLP logo≈19 GB

    We present Tongyi DeepResearch, an agentic large language model featuring 30 billion total parameters, with only 3 billion activated per token. Developed by Tongyi Lab, the model is specifically designed for long-horizon, deep information-seeking tasks.

    30.5BA3B128K ctx Apache 2.0
    11.5KSept 2025gguf
    • Same category
    • Similar size (30.5B vs 34.7B)
    • Commercial use OK
    vs
  5. ai9stars logo

    G9v3-39A5B

    ai9stars·Chat

    ≈24 GB

    G9v3-39A5B is a Mixture-of-Experts (MoE) causal language model from the AI9Stars team, with 39B total parameters and 5B activated per token.

    39BA5.4B128K ctx Apache 2.0
    868Jul 2026gguf
    • Same category
    • Similar size (39B vs 34.7B)
    • Commercial use OK
    vs
  6. Blockway logo≈22 GB

    Agens Volundr 32B is an open-weights model from Blockway, built for reasoning, coding and agent work. This is a Preview release: the architecture, format and post-training pipeline are final; continued pre-training is still running and the full v1 will replace this checkpoint.

    32.7B256K ctx Apache 2.0
    300Oct 2026gguf
    • Same category
    • Similar size (32.7B vs 34.7B)
    • Commercial use OK
    • Newer release
    vs
  7. NousResearch logo

    Hermes-4.3-36B

    NousResearch·Chat

    ≈24 GB

    Hermes 4.3 36B is a frontier, hybrid-mode reasoning model based on ByteDance Seed 36B base, made by Nous Research that is aligned to you.

    36.2B512K ctx Apache 2.0
    3.1KNov 2025gguf
    • Same category
    • Similar size (36.2B vs 34.7B)
    • Commercial use OK
    vs
  8. Fastino-Nemotron-3.5-Lightning-Finance is a 30B-parameter, 3B-active mixture-of-experts model specialized for financial reasoning, extraction, and research fine-tuned on LoRA with the Fastino Fine-Tuning Agent.

    31.6B256K ctx Apache 2.0
    618Aug 2026gguf
    • Same category
    • Similar size (31.6B vs 34.7B)
    • Commercial use OK
    vs
  9. summerAIO logo≈19 GB

    Qwen3.8-27B-Longbook27-v2 is a fine-tuned 27B model for understanding and generating long-form text such as books, papers, and lengthy reports. Version 2 improves on v1 in behavior and output quality.

    27.1B256K ctx Apache 2.0
    1.5KOct 2026
    • Same category
    • Similar size (27.1B vs 34.7B)
    • Commercial use OK
    • Newer release
    vs
  10. moonshotai logo≈32 GB

    (a) On MMLU-Pro (4k context length), Kimi Linear achieves 51.0 performance with similar speed as full attention. On RULER (128k context length), it shows Pareto-optimal performance (84.3) and 3.98x speedup.

    49.1BA3B1M ctx MIT
    49.7KOct 2025gguf
    • Same category
    • Similar size (49.1B vs 34.7B)
    • Commercial use OK
    vs
  11. XYZAILab logo

    XYZ-Aquila-mini

    XYZAILab·Chat

    ≈22 GB

    XYZ-Aquila is a family of open-weight Deep Search agents developed by XYZ AI Lab. XYZ-Aquila-mini is post-trained from Qwen3.6-35B-A3B through a bounded-exploration AI4AI pipeline: humans define the target capability, development evidence, constraints, risk boundaries, and acceptance policy, while…

    35.1BA3.9B256K ctx Apache 2.0
    333Jul 2026gguf
    • Same category
    • Similar size (35.1B vs 34.7B)
    • Commercial use OK
    vs
  12. LGAI-EXAONE logo

    EXAONE-3.5-32B-Instruct

    LGAI-EXAONE·Chat

    ≈21 GB

    We introduce EXAONE 3.5, a collection of instruction-tuned bilingual (English and Korean) generative models ranging from 2.4B to 32B parameters, developed and released by LG AI Research.

    32B32K ctx exaone
    53.5KDec 2024ollama
    • Same category
    • Similar size (32B vs 34.7B)
    vs
  13. microsoft logo

    FrogBoss-32B-2510

    microsoft·Chat

    ≈21 GB

    FrogBoss is built on the Qwen3-32B transformer architecture with a maximum context length of 64k tokens. The model uses multi-turn debugging workflows and complex code reasoning. Unlike general-purpose LLMs, FrogBoss is specialized for software engineering tasks.

    32B40K ctx MIT
    2.2KJan 2026gguf
    • Same category
    • Similar size (32B vs 34.7B)
    • Commercial use OK
    vs
  14. t-tech logo

    T-Search

    t-tech·Chat

    ≈22 GB

    We introduce T-Search, an agentic retriever built for difficult multi-step search in English and Russian. It plans and executes searches across multiple rounds with a configurable search budget.

    36BA4.7B256K ctx Apache 2.0
    316Jul 2026gguf
    • Same category
    • Similar size (36B vs 34.7B)
    • Commercial use OK
    vs
  15. ibm-granite logo≈22 GB

    Model Summary: Granite Switch 4.1 30B Preview is a modular LLM built on IBM Granite 4.1 30B with embedded adapters from the Granite Libraries collection.

    32.2B128K ctx Apache 2.0
    659May 2026
    • Same category
    • Similar size (32.2B vs 34.7B)
    • Commercial use OK
    vs
  16. Qwen logo

    Qwen3-14B

    Qwen·Chat

    ≈11 GB

    Qwen3 is the latest generation of large language models in Qwen series, offering a comprehensive suite of dense and mixture-of-experts (MoE) models.

    14.8B40K ctx Apache 2.0
    13MApr 2025ollama
    • Same category
    • Smaller (14.8B vs 34.7B)
    • Commercial use OK
    • Widely used
    vs
  17. baichuan-inc logo

    Baichuan-M2-32B

    baichuan-inc·Chat

    ≈22 GB

    This repository contains the model presented in Baichuan-M2: Scaling Medical Capability with Large Verifier System.

    32.8B128K ctx Apache 2.0
    4KAug 2025gguf
    • Same category
    • Similar size (32.8B vs 34.7B)
    • Commercial use OK
    vs
  18. nvidia logo≈22 GB

    Nemotron-Terminal is a family of models specialized for autonomous terminal interaction, fine-tuned from the Qwen3 (8B, 14B, and 32B).

    32.8B40K ctx NVIDIA Open Model License
    912Feb 2026gguf
    • Same category
    • Similar size (32.8B vs 34.7B)
    • Commercial use OK
    vs
  19. arcee-ai logo

    Trinity-Mini

    arcee-ai·Chat

    ≈16 GB

    Trinity Mini is an Arcee AI 26B MoE model with 3B active parameters. It is the medium-sized model in our new Trinity family, a series of open-weight models for enterprise and tinkerers alike.

    26.1BA3.5B128K ctx openmdw-1.1
    11.8KDec 2025gguf
    • Same category
    • Similar size (26.1B vs 34.7B)
    • Commercial use OK
    vs
  20. CohereLabs logo

    c4ai-command-r-v01

    CohereLabs·Chat

    ≈23 GB

    Cohere Labs Command-R is a research release of a 35 billion parameter highly performant generative model. Command-R is a large language model with open weights optimized for a variety of use cases including reasoning, summarization, and question answering.

    35B128K ctx CC BY-NC 4.0
    127KMar 2024gguf
    • Same category
    • Similar size (35B vs 34.7B)
    vs
  21. LiquidAI logo

    LFM2.5-24B-A2B-Base

    LiquidAI·Chat

    ≈15 GB

    LFM2.5 is a new family of hybrid models designed for on-device deployment. LFM2.5-24B-A2B is the largest model in the family: a 24B-parameter mixture of experts with only 2B active parameters.

    24BA2B128K ctx lfm1.0
    2.5KOct 2026
    • Same category
    • Similar size (24B vs 34.7B)
    • Newer release
    vs
  22. apple logo≈19 GB

    This model is an example of the Simple Self-Distillation (SimpleSD) method that improves code generation by fine-tuning a language model on its own sampled outputs—without rewards, verifiers, teacher models, or reinforcement learning. Please see the paper below for more information.

    30.5BA3.4B256K ctx Apple Machine Learning Research License
    1.2KMar 2026gguf
    • Same category
    • Similar size (30.5B vs 34.7B)
    vs
  23. LGAI-EXAONE logo

    EXAONE-4.0.1-32B

    LGAI-EXAONE·Chat

    ≈21 GB

    ✈️ Try on FriendliAI (licensed under commercial purposes) 📢 EXAONE 4.0 is officially supported by HuggingFace transformers! Please check out the guide below

    32B128K ctx exaone
    4.1KJul 2025gguf
    • Same category
    • Similar size (32B vs 34.7B)
    vs
  24. mindlab-research logo

    Macaron-V1-Tall

    mindlab-research·Chat

    ≈22 GB

    Macaron-V1-Tall is a Mixture of LoRA (MoL) model in the Macaron-V1 family, built on Qwen3.6-35B-A3B for personal intelligence, tool use, coding workflows, and code-native Generative UI.

    36BA4.7B256K ctx MIT
    183Jul 2026gguf
    • Same category
    • Similar size (36B vs 34.7B)
    • Commercial use OK
    vs