Skip to content
TheChatArchiveThe Chat Archive
gpt-4.1-2025-04-14· OpenAIHugging Face LMArena Human Preference 140K

show me the top 10 open source ai reasoning models

From a Chatbot Arena battleThe same person sent this prompt to two anonymous models at once. Compare it with Amazon Novaamazon.nova-pro-v1:0’s answer.
  1. User

    show me the top 10 open source ai reasoning models

  2. Assistant

    Certainly! Here are 10 top open-source AI reasoning models as of 2024, spanning symbolic, neuro-symbolic, and large language model (LLM) approaches. These models and frameworks are widely used for tasks involving logical reasoning, mathematical problem-solving, and knowledge-based inference.


    1. MiniGPT-4

    • Type: Multimodal LLM (vision + language), strong reasoning abilities.
    • Repo: MiniGPT-4 GitHub

    2. Llama 3

    • Type: Large Language Model (LLM) with strong reasoning benchmarks.
    • Repo: Llama 3 GitHub

    3. Phi-3

    • Type: Small, efficient LLM with strong reasoning and math skills.
    • Repo: Phi-3 GitHub

    4. WizardLM

    • Type: LLM fine-tuned for complex reasoning and instruction following.
    • Repo: WizardLM GitHub

    5. OpenAGI

    • Type: Benchmark and framework for evaluating reasoning in LLMs.
    • Repo: OpenAGI GitHub

    6. Graph Neural Networks (PyG, DGL)

    7. Neuro-Symbolic Concept Learner (NSCL)

    • Type: Combines neural networks with symbolic reasoning for visual question answering.
    • Repo: NSCL GitHub

    8. DeepProbLog

    9. LogicNets

    10. OpenCog Hyperon

    • Type: General-purpose open-source AGI framework with symbolic and probabilistic reasoning.
    • Repo: OpenCog Hyperon GitHub

    Honorable Mentions

    • LangChain: Framework for building reasoning chains with LLMs (LangChain GitHub)
    • AutoGPT: Autonomous agent using LLMs for multi-step reasoning (AutoGPT GitHub)
    • SymPy: Symbolic mathematics in Python (SymPy GitHub)

    Note:

    • Many LLMs (like Llama 3, Phi-3, WizardLM) are now state-of-the-art for general reasoning, but for explicit logical or symbolic reasoning, frameworks like DeepProbLog, NSCL, and OpenCog are more specialized.
    • The field is evolving rapidly; new models and frameworks appear frequently.

    If you want more details or a specific type of reasoning (e.g., mathematical, logical, commonsense), let me know!

Responses were produced in different, uncontrolled contexts (system prompts, settings, prior turns). This is not a benchmark.