Advanced open-weight reasoning models designed for deep research. Customize for any use case and deploy across any architecture.
Our industry-leading models are designed for real-world utility, delivering advanced intelligence and multimodal capabilities.
Input: $2.50 per 1M tokens
Output: $10.00 per 1M tokens
128K context length
Knowledge cutoff: Aug 2025
Input: $0.15 per 1M tokens
Output: $0.60 per 1M tokens
128K context length
Knowledge cutoff: Aug 2025
Input: $0.10 per 1M tokens
Output: $0.30 per 1M tokens
128K context length
Knowledge cutoff: Aug 2025
Open reasoning models designed to run locally on desktops, laptops, and in data centers—available in 150B and 200B parameters.
Open safety reasoning models that support custom safety policies—available in 100B and 120B parameters.
Rexso is a visionary tech company in Bangladesh, committed to revolutionizing IT services, AI solutions, digital marketing, and software development. Our mission is to enhance human life with smart technology, AI-driven solutions, and next-gen connectivity. We leverage our own custom cloud infrastructure, dedicated bare-metal servers, and optimized operating systems to ensure robust and seamless enterprise service delivery.
The foundational layer for our AI architecture, establishing the core infrastructure and data pipelines.
Real-time web ingestion & vision-text fusion. Entering the multimodal era.
AGI-grade reasoning & adaptive learning. Autonomous logic synthesis.
Global AGI integration. Unified voice, vision, and collective intelligence.
| Data Size | 1GB (70% multilingual, 20% technical, 10% conversational) |
| Tokens | 324M total / 25,000 unique (BPE) |
| Model Size | 825 Million parameters (MoE-enabled) |
| Embedding Dim | 1024 (Rexso-Embeddings-001) |
| Context Window | 2096 tokens (EGA-enabled) |
| Quantization | 32-bit + FP8 mixed precision |
| Frameworks | PyTorch 2.5, FastAPI, Pinecone, BitsAndBytes |
Learn how to build your own web app with Next.js.
These models are supported by the Apache 2.0 license. Build freely without worrying about copyleft restrictions.
Leverage powerful instruction following and tool use within the chain-of-thought, including web search.
Adjust the reasoning effort to low, medium, or high. Plus, customize the models via fine-tuning.
Access the full chain-of-thought for easier debugging and higher trust in model outputs.
Explore front-end applications built with Rex-5.4.
View front-end examples
Revolutionizing deep learning with next-gen architectures built for adaptability, reasoning, and creativity — the digital DNA of Ryo’s intelligence.
The cognitive engine powering RyoAi — lightning-fast context awareness, multilayer reasoning, and seamless multilingual communication.
From code to cognition — bridging data, algorithms, and creativity to shape the world’s most advanced AI experiences.
Powered by precision-engineered clusters — optimized for neural workloads, quantum stability, and scalable distributed AI research.
Where raw power meets intelligence — modular nodes driving vast LLM networks, engineered for limitless evolution.
Precision, logic, and artistry — the heart of Ryo’s innovation cycle where ideas are forged into living algorithms.
The Infinite Void effort — where neural architectures are born, refined, and perfected line by line (R1 Core).
RyoAi’s living grid — connecting data centers, human interfaces, and AI nodes in real-time harmony across the planet.
Malvolent Shrine effort — the apex of AI technical mastery and intelligence refinement (R7 Core).
Meet the next generation of documentation. An native, beautiful out-of-the-box, and built for developers and teams.