IT Brief New Zealand - Technology news for CIOs & IT decision-makers
Story image
Anthropic unveils Claude 3 AI models with boosted intelligence & speed
Mon, 18th Mar 2024

Anthropic has announced the launch of Claude 3, the next generation of its vision-enabled AI models. The latest offerings from the artificial intelligence company, Claude 3 Opus, Claude 3 Sonnet, and Claude 3 Haiku, are designed for various applications and strike an effective balance between intelligence, speed, and cost. Each model boasts improvements in several domains, ranging from content creation to multilingual capabilities and coding.

The highlight of the new model series is Claude 3 Opus, touted as the most capable model globally on a wide spectrum of benchmarks. Capable of top-tier performance on highly complex tasks, it sets a significant "new market standard," demonstrating the potential of generative AI.

Specifically designed for high-volume work, Claude 3 Sonnet is two times faster than its predecessors, displaying greater intelligence. Making the most of speed and affordability, Claude 3 Haiku is a versatile solution for various enterprise applications, positioning itself as the fastest and most affordable model in its class.

Beyond improved speed, Claude 3 models excel in content creation, data analysis and forecasting, code generation, and language translation, among other features. Enhancements in vision capabilities allow these models to process a myriad of visuals, such as photos, charts, graphs, and technical diagrams. Coding proficiency has also seen an upgrade, with Claude 3 capable of answering user queries about a codebase or explaining a piece of code.

With the promise of enhanced fluency in non-English languages, including Spanish, French, and Japanese, these new models open up possibilities for translation services and international customer support.

The new models present a two-fold increase in task completion speed compared to previous iterations. Particularly in live chats, data extraction tasks, and auto-completion scenarios, these models deliver near-instant results. Haiku, being the fastest and most cost-effective model, can process data-dense research papers, inclusive of charts and graphs, in less than three seconds.

In addition to superior speed and intelligence, the Claude 3 models display a more nuanced understanding of requests and improved accuracy for businesses reliant on model outputs for customer service. The Claude 3 Opus model, for instance, demonstrates a notable improvement in its ability to recall information accurately from a large data corpus.

The models' responsible design ensures their trustworthiness, and a concerted effort has been made to reduce biases. Despite advancements in key areas of knowledge, these models maintain an AI Safety Level 2 (ASL-2) according to Anthropic's Responsible Scaling Policy.

Anthropics says that their efforts in pushing AI boundaries, coupled with stringent safety guardrails, will steer AI development towards positive societal outcomes.