Search Results for "reasoning" - Page 2
Sort By:
Showing 271 open source projects for "reasoning"
View related business solutions-
Ship Agents FasterGemini Enterprise Agent Platform lets you rapidly build, scale, govern and optimize production-ready agents grounded in your organization's data. The platform enables developers to build custom or pre-built agents for virtually any use case. New customers get $300 in free credits.
-
Demo Series - Small Business Backup By VeeamWatch this on-demand demo series and learn how to protect your Microsoft 365 data with clear, simple, actionable steps that are easy to implement for businesses of all sizes.
-
1
GLM-4.5V
GLM-4.6V/4.5V/4.1V-Thinking, towards versatile multimodal reasoning
GLM-4.5V is the preceding iteration in the GLM-V series that laid much of the groundwork for general multimodal reasoning and vision-language understanding. It embodies the design philosophy of mixing visual and textual modalities into a unified model capable of general-purpose reasoning, content understanding, and generation, while already supporting a wide variety of tasks: from image captioning and visual question answering to content recognition, GUI-based agents, video understanding, and long-document interpretation. ... -
2
PageIndex
Document Index for Vectorless, Reasoning-based RAG
...Rather than chunking text and embedding it into a vector database, PageIndex constructs a tree-structured index — similar to a detailed, AI-enhanced table of contents — that a large language model can traverse to locate the most relevant sections of long documents. This reasoning-driven retrieval aligns more naturally with how humans explore complex texts, improving relevance and traceability, especially in professional domains like financial reports, legal contracts, and technical manuals. The project includes example notebooks, scripts for tree generation and search, and support for multiple document formats including PDF and markdown, with tools designed to preserve context and semantic boundaries. -
3
VibeThinker
Diversity-driven optimization and large-model reasoning ability
VibeThinker is a compact but high-capability open-source language model released by WeiboAI (Sina AI Lab). It contains about 1.5 billion parameters, far smaller than many “frontier” models, yet it is explicitly optimized for reasoning, mathematics, and code generation tasks rather than general open-domain chat. The innovation lies in its training methodology: the team uses what they call the Spectrum-to-Signal Principle (SSP), where a first stage emphasizes diversity of reasoning paths (the “spectrum” phase) and a second stage uses reinforcement techniques (the “signal” phase) to refine toward correctness and strong reasoning. -
4
Qwen2.5-Math
A series of math-specific large language models of our Qwen2 series
...It includes base models (1.5B / 7B / 72B parameters), instruction-tuned versions, and a reward model (RM) to improve alignment. Unlike its predecessor Qwen2-Math, Qwen2.5-Math supports both Chain-of-Thought (CoT) reasoning and Tool-Integrated Reasoning (TIR) for solving math problems, and works in both Chinese and English. It is optimized for solving mathematical benchmarks and exams; the 72B-Instruct model achieves state-of-the-art results among open source models on many English and Chinese math tasks. -
MongoDB Atlas runs apps anywhereMongoDB Atlas gives you the freedom to build and run modern applications anywhere—across AWS, Azure, and Google Cloud. With global availability in over 115 regions, Atlas lets you deploy close to your users, meet compliance needs, and scale with confidence across any geography.
-
5
GLM-4.5
GLM-4.5: Open-source LLM for intelligent agents by Z.ai
...The models support FP8 and BF16 precision, and can handle very large context windows of up to 128K tokens. Flexible inference is supported through frameworks like vLLM and SGLang with tool-call and reasoning parsers included. -
6
MiroThinker
MiroThinker is an open source deep research agent
MiroThinker is an open-source deep research AI agent designed to perform complex reasoning, information gathering, and predictive analysis tasks. The system focuses on enabling long-horizon research workflows by allowing the agent to interact repeatedly with external tools, search systems, and data sources while refining its reasoning through iterative steps. Rather than simply generating responses from a single prompt, the agent performs structured multi-step reasoning processes that involve searching for information, analyzing evidence, and synthesizing conclusions. ... -
7
R-KV
Redundancy-aware KV Cache Compression for Reasoning Models
...Modern transformer models rely heavily on KV caches during autoregressive decoding, which store intermediate attention states to accelerate generation. However, these caches can consume large amounts of memory, especially in reasoning-oriented models with long context windows. R-KV introduces a method for compressing the KV cache during decoding, allowing models to maintain reasoning performance while reducing memory consumption and computational overhead. The approach focuses on identifying which attention heads and cache components are most important for maintaining reasoning quality, allowing less critical information to be compressed or discarded. ... -
8
AReal
Lightning-Fast RL for LLM Reasoning and Agents. Made Simple & Flexible
AReaL is an open source, fully asynchronous reinforcement learning training system. AReal is designed for large reasoning and agentic models. It works with models that perform reasoning over multiple steps, agents interacting with environments. It is developed by the AReaL Team at Ant Group (inclusionAI) and builds upon the ReaLHF project. Release of training details, datasets, and models for reproducibility. It is intended to facilitate reproducible RL training on reasoning / agentic tasks, supporting scaling from single nodes to large GPU clusters. ... -
9
MiniMax-M1
Open-weight, large-scale hybrid-attention reasoning model
MiniMax-M1 is presented as the world’s first open-weight, large-scale hybrid-attention reasoning model, designed to push the frontier of long-context, tool-using, and deeply “thinking” language models. It is built on the MiniMax-Text-01 foundation and keeps the same massive parameter budget, but reworks the attention and training setup for better reasoning and test-time compute scaling. Architecturally, it combines Mixture-of-Experts layers with lightning attention, enabling the model to support a native context length of 1 million tokens while using far fewer FLOPs than comparable reasoning models for very long generations. ... -
Custom VMs From 1 to 96 vCPUs With 99.95% UptimeLive migration and automatic failover keep workloads online through maintenance. One free e2-micro VM every month.
-
10
DriveLM
Driving with Graph Visual Question Answering
DriveLM is a research-oriented framework and dataset designed to explore how vision-language models can be integrated into autonomous driving systems. The project introduces a new paradigm called graph visual question answering that structures reasoning about driving scenes through interconnected tasks such as perception, prediction, planning, and motion control. Instead of treating autonomous driving as a purely sensor-driven pipeline, DriveLM frames it as a reasoning problem where models answer structured questions about the environment to guide decision making. The system includes DriveLM-Data, a dataset built on driving environments such as nuScenes and CARLA, where human-written reasoning steps connect different layers of driving tasks. ... -
11
ReCall
Learning to Reason with Search for LLMs via Reinforcement Learning
...The framework uses reinforcement learning to train models to perform these tool calls effectively while solving multi-step reasoning tasks. -
12
NagaAgent
A simple yet powerful agent framework for personal assistants
NagaAgent is an experimental framework for building interactive virtual agents capable of autonomous reasoning, dialog, and task execution using components that mirror human cognitive patterns. It provides abstractions for representing goals, context, and state so that agents can plan sequences of actions, evaluate outcomes, and adjust behavior over time. The project includes mechanisms for semantic memory, reasoning pipelines, and integration points with external data sources and language models so that agents can interpret natural language instructions and produce coherent multi-step outputs. ... -
13
gpt-oss
gpt-oss-120b and gpt-oss-20b are two open-weight language models
gpt-oss is OpenAI’s open-weight family of large language models designed for powerful reasoning, agentic workflows, and versatile developer use cases. The series includes two main models: gpt-oss-120b, a 117-billion parameter model optimized for general-purpose, high-reasoning tasks that can run on a single H100 GPU, and gpt-oss-20b, a lighter 21-billion parameter model ideal for low-latency or specialized applications on smaller hardware. -
14
Bonsai 27B
Run Bonsai (1-bit) and Ternary-Bonsai language models locally
...The models can run on macOS, Linux, and Windows through CPU, Metal, CUDA, Vulkan, ROCm, llama.cpp, or MLX backends. Its 27B models process text, images, screenshots, and PDFs while supporting reasoning and long-context conversations. They also provide OpenAI-compatible tool calling and optional MCP server integration for agentic workflows. Automated setup scripts install dependencies, download models, obtain binaries, and configure interactive interfaces. Users can access the models through command-line prompts, a local chat server, or Open WebUI. -
15
QwenPaw
A personal AI assistant, easy to install
...The project combines conversational AI, memory systems, tool usage, workflow automation, and multimodal interaction into a unified assistant environment designed for daily productivity and experimentation. It supports structured reasoning, autonomous task execution, and integration with external tools and APIs, allowing the assistant to perform actions beyond standard chatbot conversations. QwenPaw emphasizes extensibility through modular components that developers can customize for research, personal assistants, automation systems, or enterprise AI workflows. The architecture integrates agent collaboration, planning systems, and long-term contextual memory to create more persistent and adaptive interactions. -
16
Synthetic Data Kit
Tool for generating high quality Synthetic datasets
...It supports generation of rationales/chain-of-thought variants, configurable sampling, and guardrails so outputs meet format constraints and quality checks. Examples and guides show how to target task-specific behaviors like tool use or step-by-step reasoning, then save directly into training-ready files. -
17
RecursiveMAS
Offical Implementation for "Recursive Multi-Agent Systems"
RecursiveMAS is an advanced multi-agent AI framework that introduces a recursive collaboration mechanism to improve reasoning and problem-solving across multiple agents. Instead of treating agents as independent units exchanging text outputs, it connects them through a shared latent computation loop, allowing internal “thought states” to be passed and refined iteratively. This recursive structure enables agents to build on each other’s intermediate reasoning, leading to deeper and more coherent solutions. ... -
18
OmAgent
Build multimodal language agents for fast prototype and production
...Instead of forcing developers to implement complex orchestration logic manually, the system manages task scheduling, worker coordination, and node optimization behind the scenes. Its architecture uses a graph-based workflow engine where tasks are represented as nodes in a directed workflow, enabling modular composition of complex reasoning pipelines. The framework also includes support for various reasoning strategies commonly used in language agents, such as chain-of-thought prompting, self-consistency reasoning, and ReAct-style decision loops. -
19
Mathematics Dataset
This dataset code generates mathematical question and answer pairs
...The dataset enables models to learn mathematical problem-solving through examples that involve both numeric and symbolic reasoning. Version 1.0 includes over 2 million examples per category, with training splits labeled as “easy,” “medium,” and “hard,” supporting curriculum-based learning strategies. The data can be accessed via PyPI or generated locally using provided Python scripts, with outputs formatted for direct use in training or evaluation pipelines. -
20
Controllable-RAG-Agent
This repository provides an advanced RAG
...Instead of relying solely on simple semantic search, it builds a deterministic control graph that acts as the “brain” of the agent, orchestrating planning, retrieval, reasoning, and verification across many steps. The pipeline ingests PDFs, splits them into chapters, cleans and preprocesses text, then constructs vector stores for fine-grained chunks, chapter summaries, and book quotes to support nuanced queries. At query time, it anonymizes entities, creates a high-level plan, de-anonymizes and expands that plan into concrete retrieval or reasoning tasks, and executes them in sequence while continuously revising the plan. ... -
21
MobileLLM
MobileLLM Optimizing Sub-billion Parameter Language Models
MobileLLM is a lightweight large language model (LLM) framework developed by Facebook Research, optimized for on-device deployment where computational and memory efficiency are critical. Introduced in the ICML 2024 paper “MobileLLM: Optimizing Sub-billion Parameter Language Models for On-Device Use Cases”, it focuses on delivering strong reasoning and generalization capabilities in models under one billion parameters. The framework integrates several architectural innovations—SwiGLU activation, deep and thin network design, embedding sharing, and grouped-query attention (GQA)—to achieve a superior trade-off between model size, inference speed, and accuracy. MobileLLM demonstrates remarkable performance, with the 125M and 350M variants outperforming previous state-of-the-art models of the same scale by up to 4.3% on zero-shot commonsense reasoning tasks. -
22
GenericAgent
Self-evolving autonomous agent framework
The GenericAgent project is a flexible framework for building autonomous AI agents that can operate across diverse tasks and environments. It is designed around modularity, allowing developers to define agents with interchangeable components such as tools, memory systems, and reasoning strategies. The architecture emphasizes generality, enabling the same agent framework to be adapted for different domains including coding, research, and task automation. It integrates with modern language models to provide planning, execution, and iterative reasoning capabilities, making it suitable for complex workflows. The project also focuses on extensibility, allowing developers to plug in custom tools or APIs and tailor agent behavior to specific use cases. ... -
23
Agno
Lightweight framework for building Agents with memory, knowledge, etc.
Agno is a modular, open-source artificial general intelligence (AGI) research platform that allows developers to build, evaluate, and experiment with cognitive architectures in a composable way. It provides a flexible framework for modeling reasoning, memory, decision-making, and planning, aimed at long-term AI research beyond narrow learning. Agno embraces multi-agent environments and symbolic reasoning as part of its core design, enabling experiments with structured knowledge, goal-oriented behaviors, and meta-learning. It’s designed for researchers seeking an extensible platform to explore AGI components without being tied to black-box models. -
24
OpenManus
Open-source AI agent framework
...It is built to be flexible and extensible, enabling customization of agent behaviors, tools, and reasoning strategies. Overall, OpenManus provides a foundation for creating more capable, autonomous AI systems that can handle dynamic and goal-driven tasks. -
25
bert4torch
An elegent pytorch implement of transformers
An elegant PyTorch implement of transformers.