Skip to main content

China Launches First Open-Source Oceanic AI Model OceanGPT

China Unveils Open-Source Oceanic AI Model OceanGPT

On June 26, OceanGPT (Cangyuan), China's first open-source large model specialized for marine applications, was officially launched in Hangzhou. Developed primarily by Zhejiang University's National Key Laboratory of Marine Precision Sensing Technology, this breakthrough aims to revolutionize marine scientific research and practical applications.

Advanced Capabilities for Marine Research

OceanGPT demonstrates multi-modal data interpretation, processing sonar images and marine observation maps with natural language outputs. Its innovative "slow thinking" reasoning mechanism combines domain-specific knowledge to minimize errors—addressing the hallucination issues common in general-purpose AI models.

Image

Image source note: The image is AI-generated, and the image licensing service provider is Midjourney

Transforming Underwater Robotics

The model extends beyond Q&A functionality to enable natural language-controlled marine robotics. Professor Chen Huajun, lead researcher at Zhejiang University, explained that OceanGPT can generate robot control code from simple instructions, democratizing access to complex underwater operations for non-specialists.

Key Technological Integration

  • MCP large model protocol integration enables multi-robot collaboration
  • Field-tested on Zhejiang University's Haiying underwater robots
  • Reduces code generation time from hours to seconds

Development Timeline and Architecture

Initiated in 2023, OceanGPT has evolved through multiple versions:

  1. OceanGPT-basic: Marine knowledge Q&A
  2. OceanGPT-omni: Multi-modal data processing
  3. OceanGPT-coder: Underwater robot programming

The model employs a mixture of experts architecture, where specialized sub-models collaborate to expand marine applications.

Key Points:

  • First open-source oceanic AI model developed in China
  • Reduces error rates through domain-specific "slow thinking"
  • Enables natural language control of underwater robotics
  • Integrates MCP protocol for multi-agent collaboration
  • Cuts robot programming time from hours to seconds

Enjoyed this article?

Subscribe to our newsletter for the latest AI news, product reviews, and project recommendations delivered to your inbox weekly.

Weekly digestFree foreverUnsubscribe anytime

Related Articles

Alibaba's 'Happy Oyster' Opens New Chapter in Interactive AI World
News

Alibaba's 'Happy Oyster' Opens New Chapter in Interactive AI World

Alibaba's ATH team has unveiled Happy Oyster, an open-world AI model that brings real-time interaction to virtual environments. Building on the success of HappyHorse - which recently topped global video editing rankings - this innovative tool marks the company's ambitious move into dynamic AI worlds. Developers and creators can now apply for early access to what could become a game-changer for virtual simulations and digital design.

April 16, 2026
Artificial IntelligenceInteractive TechnologyAlibaba Innovation
Apple's Siri Team Gets Intensive AI Training to Close the Gap
News

Apple's Siri Team Gets Intensive AI Training to Close the Gap

Apple is putting its Siri engineers through an intensive AI bootcamp, signaling a major push to transform its voice assistant into a true AI companion. The program focuses on practical skills like prompt engineering and privacy-focused AI deployment. This comes as Apple seeks to address criticisms about falling behind in the AI race while maintaining its signature focus on user privacy.

April 16, 2026
AppleSiriArtificial Intelligence
Anthropic's Secret AI Model Mythos Showcased to Trump Team
News

Anthropic's Secret AI Model Mythos Showcased to Trump Team

Anthropic co-founder Jack Clark revealed at the Semafor summit that his company demonstrated its unreleased AI model Mythos to Trump administration officials, citing its advanced cybersecurity capabilities. Despite an ongoing legal battle with the Pentagon over military AI use, Clark emphasized the importance of government-tech collaboration. The revelation comes as major banks reportedly test the powerful new system, while Clark offers surprising optimism about AI's employment impact compared to his CEO's dire predictions.

April 15, 2026
Artificial IntelligenceCybersecurityGovernment Tech
News

Investors Shift Focus as Anthropic Surges, Leaving OpenAI Valuation in Question

The AI investment landscape is shifting dramatically as Anthropic's explosive growth challenges OpenAI's dominance. While OpenAI boasts an $852 billion valuation, Anthropic's revenue tripled to $30 billion in just three months, fueled by demand for its coding tools. Secondary market activity shows investors favoring Anthropic's stock, trading OpenAI shares at a discount. The rivalry echoes past tech battles where early leaders didn't always finish first.

April 15, 2026
AI InvestmentTech ValuationArtificial Intelligence
News

NVIDIA Unveils Open-Source Quantum AI Breakthrough, Sparking Market Rally

NVIDIA has made waves in the tech world by releasing Ising, the first open-source quantum AI model series. This groundbreaking development addresses critical challenges in quantum computing, offering researchers tools to build more reliable quantum processors. The announcement sent quantum tech stocks soaring, with some Chinese companies seeing massive fund inflows. The Ising models - one for calibration, another for error correction - promise speeds 2.5 times faster than current standards with a 300% accuracy boost.

April 15, 2026
Quantum ComputingArtificial IntelligenceNVIDIA
News

DeepMind CEO Predicts AGI Within Five Years: A Revolution Unlike Any Before

DeepMind CEO Demis Hassabis has made bold predictions about artificial intelligence's future, suggesting AGI could arrive within five years. He describes this shift as a "tenfold industrial revolution happening ten times faster" than historical changes. Hassabis also warns about widening gaps between top AI companies and the patchy nature of current AI systems. The interview reveals how the rules of AI development are changing, with innovation becoming more crucial than raw computing power.

April 14, 2026
AGIDeepMindAI Future