更新日時：　投稿日時：2023-10-26

The pace of innovation in artificial intelligence is breathtaking. Every few months, we cross a new threshold of what's possible. Today, we at Project Singularity are thrilled to take the next significant step forward with the release of Nexus-4, our most advanced and capable foundation model.

Nexus-4 isn't just an incremental update; it's a fundamental architectural shift designed for a world where intelligence needs to be seamless, contextual, and multimodal.

What's New in Nexus-4?

We focused on four key pillars during the development of Nexus-4: multimodality, context, reasoning, and efficiency.

• True Multimodality: Nexus-4 was designed from the ground up to understand and process information from different formats simultaneously. It can natively interpret a mix of text, images, and audio in a single prompt. You can now ask it to "describe the mood of this image and suggest a short audio clip that matches it" without any complex pre-processing.

• A Massive Context Window: The model features a 1 million token context window. This allows for the processing of entire books, extensive codebases, or hours of transcribed conversation in a single pass. This unlocks new possibilities for deep document analysis, complex code refactoring, and maintaining long-term conversational memory.

• Enhanced Reasoning Capabilities: We have significantly improved the model's logical reasoning and problem-solving skills. Nexus-4 excels at multi-step mathematical problems, complex scientific analysis, and generating high-quality, bug-free code.

• Blazing-Fast and Efficient: Despite its immense power, Nexus-4 is our fastest and most cost-effective flagship model to date. We've optimized the architecture to deliver lower latency and require less computational power, making it accessible for a wider range of applications and developers.

Nexus-4 doesn't just process information; it understands context across different formats, leading to richer and more accurate insights.

What Can You Build with Nexus-4?

We believe Nexus-4 will empower developers and businesses to create applications that were previously science fiction. Here are just a few ideas:

• Intelligent Support Agents: Build customer service bots that can understand a user's typed problem, look at a screenshot they've uploaded, and listen to a voice note explaining their frustration—all in one interaction.
• Creative Co-Pilots: Generate entire marketing campaigns, including ad copy, accompanying images, and scripts for short promotional videos, all from a simple creative brief.
• Advanced Data Analysis: Feed Nexus-4 dense financial reports, charts, and market commentary to get a comprehensive, human-readable summary of the current economic landscape.
• Accessible Tools: Create applications that can describe the world for visually impaired users in rich detail or transcribe and summarize complex meetings for those with hearing impairments.

Built with Responsibility in Mind

With great power comes great responsibility. Nexus-4 has undergone the most extensive safety testing of any of our models. We have worked with a diverse group of red-teamers and external experts to mitigate biases, reduce harmful outputs, and ensure the model aligns with our ethical AI principles. We are committed to providing developers with the tools and guidance needed to build safe and beneficial applications.

Get Started Today

We are incredibly excited to see what you will build with Nexus-4. The model is available starting today via our API.

• Read the API Documentation to learn about the new multimodal endpoints.
• Explore our Quickstart Guide to make your first call in minutes.
• Join our Community Forum to share your projects and get support from our team.

The future is multimodal, and it starts now. Let's build it together.