Photo: PatrickAssale / Shutterstock.com
AI company Anthropic has launched a new AI model, the Claude 3.5 Sonnet, which it claims is its best. Alongside the new model, Anthropic is also launching Artifacts — a workspace where users can edit and add to content generated by Anthropic’s models.
Claude 3.5 Sonnet is free on Claud.ai and the Claude iOS app, although with limited rate limits. Claude Pro and Team plan subscriptions offer access to the model at “significantly higher rate limits” for those who want more.
Claude 3.5 Sonnet can analyse text and images and generate text. It also outperforms Claude 3 Sonnet, the model it’s replacing, and even managed to beat Anthropic’s previous flagship model, the Claude 3 Opus. Although Anthropic boasts that its new model is setting “new industry benchmarks,” it’s only barely beating rival models, especially OpenAI’s GPT-40, on several benchmarks that Anthropic tested it against.
The company claims that its new model can better understand nuanced and complex instructions and concepts like humour. It also greatly improves vision, interpreting charts and graphs and transcribing text from imperfect images, including those with distortions and visual artefacts.
Many of these improvements have come because of the new model’s training data. However, the company hasn’t revealed what data it used in the training process. This could be because of the very high competition in the AI industry at the moment or to avoid getting drowned in legal troubles, especially considering challenges regarding fair use.
In the News: Amazon’s AI-powered Alexa to come with a monthly price tag