Anthropic Launches Claude3.5 Sonnet with Superior Performance

TapTechNews June 21, Anthropic today (June 21) launched Claude3.5 Sonnet, This is the first model version of the Claude3.5 series and outperforms OpenAI's GPT-4o and Google's Gemini1.5Pro in various evaluations.

Claude3.5 Sonnet iteratively upgrades on the basis of the previous generation, with faster performance and stronger coding, visual and natural language understanding capabilities.

Claude3.5 Sonnet is a medium-sized model, between the small Haiku and the high-end Opus, but Anthropic claims that according to internal benchmark tests, the Sonnet's performance even exceeds that of the top Opus.

According to Anthropic, Claude3.5 Sonnet processes input twice as fast as Opus. The company said that in coding challenges, Sonnet outperformed the previous Opus model and fixed 64% of the code errors, while Opus only 38%.

Benchmark test results also show that Sonnet beats rivals in 7 of the 9 overall categories and 4 of the 5 visual tasks.

TapTechNews translated the official press release as follows:

Claude3.5 Sonnet is our strongest visual model to date, surpassing Claude3 Opus in standard visual benchmark tests, and these step-like improvements are most apparent in tasks that require visual reasoning such as interpreting charts.

Claude3.5 Sonnet can also accurately transcribe text from imperfect images, which is a core competency in retail, logistics and financial services, where AI may gain more insights from images, graphics or illustrations than from text alone.

To ensure safety, Anthropic submitted Sonnet to the UK and US AI Safety Institutes for external evaluation. They found that after improvement, the model is still at the ASL2 level. Anthropic has also updated the assistant with child safety expertise to better prevent potential harm.