Reka unveils Yasa-1: A multimodal AI competitor to ChatGPT

Reka's Yasa 1 enters the ring, challenging ChatGPT, and promising diverse skills in understanding and generating content across various modes.

Reka unveils Yasa 1
Reka unveils Yasa 1

Highlights

  • Reka's Yasa-1 is not an ordinary AI— it understands not just words but also images, audio, and short videos, making it a versatile assistant
  • Reka envisions Yasa-1 as a game-changer, customisable for various uses

Reka, an AI startup founded by researchers from DeepMind, Google, Baidu, and Meta, has introduced Yasa-1, a powerful multimodal AI assistant set to rival ChatGPT's dominance. Yasa-1 goes beyond text, understanding images, short videos, and audio snippets.

Yasa-1's versatility

Yasa-1 offers more than just text comprehension. This AI assistant can be customised on private datasets, empowering enterprises to create new experiences across various use cases. Supporting 20 languages, it provides answers with internet context, processes long documents, and even executes code.

Direct competition with ChatGPT

Yasa-1 enters the ring as a direct competitor to ChatGPT, which recently received its multimodal upgrade. Both aim to provide diverse AI capabilities, and Yasa-1's unique features could reshape the AI landscape.

Multimodal capabilities

Available through APIs and docker containers, Yasa-1 leverages a unified model to understand not just words, but also images, audio, and short videos. This multimodal understanding allows users to combine text prompts with multimedia, enabling more specific and detailed responses.

Use cases and limitations

Yasa-1's capabilities extend to tasks like generating social media posts from product images or detecting sounds. However, it's highlighted that for the best experience, audio or video prompts should be under one minute. The technology is still evolving, with the acknowledgment that the model may 'hallucinate’' and should not be solely relied upon for critical advice.

Additional features

Yasa-1 boasts support for 20 languages, long document processing, and the unique ability to execute code (exclusive to on-premise deployments). It can perform arithmetic operations, analyse spreadsheets, and create visualisations. The integration of real-time web content into answers further enhances its utility.

Looking ahead

In the coming weeks, Reka plans to expand Yasa-1's access to more enterprises while actively addressing and improving its limitations. The company envisions Yasa-1 as a generative agent with multimodal capabilities, a crucial step towards its long-term mission of creating beneficial AI solutions.

Competitive landscape

While Reka is relatively new, having emerged from stealth just three months ago with significant funding, it faces competition from established players like OpenAI and Anthropic. The AI race is dynamic, and Reka's unique approach positions it as a formidable contender with the potential to reshape the future of AI.

Yasa-1 is a new, versatile player in AI, aiming to team up with humans to tackle big challenges. Exciting updates are expected soon.