Anthropic, the Amazon-supported competitor to OpenAI, has launched a significant upgrade to its flagship AI model, introducing Claude 3.5 Sonnet. This mid-tier model enhances upon the performance of Claude 3, as the company asserts it “raises the industry bar for intelligence.” Claude 3.5 Sonnet outshines rival models such as OpenAI’s latest GPT-4o and Google’s Gemini 1.5 Pro.
One of the notable aspects of the Claude 3.5 Sonnet is its cost-effectiveness. Operating at just $3 per million input tokens and $15 for each million output tokens, it is designed to be less expensive than its predecessors. Its extensive token context window allows it to process up to 200,000 tokens in a single prompt, making the maximum cost for input only $0.60.
Available through the Claude web application and iOS app, Claude 3.5 Sonnet offers enhanced access for Claude Pro and Team subscribers, providing higher rate limits than those available to free users. Enterprises can also take advantage of the model via Anthropic’s API, along with third-party platforms like Amazon Bedrock and Google Cloud’s Vertex AI.
Amazon, a major investor with $4 billion in Anthropic, describes Claude 3.5 Sonnet as “the ideal model for complex tasks,” such as context-sensitive customer support, orchestrating multi-step workflows, translating code, and developing revenue-generating applications.
This launch marks the beginning of a new series of upgraded Claude models, with additional versions, including the compact Haiku and large Opus, set to debut later this year. Claude 3.5 Sonnet demonstrates substantial improvements in content generation capabilities, producing nuanced and tonal writing twice as rapidly as the largest previous model, Opus. It boasts enhanced visual interpretation skills, allowing it to analyze charts and graphs, as well as transcribe text from distorted images.
In performance benchmarks, Claude 3.5 Sonnet achieved impressive scores, notably on MMLU for general knowledge and the coding evaluation HumanEval. It surpassed the earlier Claude 3 Opus model and also outperformed GPT-4o and Gemini 1.5 Pro on various assessments including MathVista, AI2D, and ANLS Score.
According to Anthropic, when provided with proper instructions and resources, Claude 3.5 Sonnet exhibits advanced reasoning and troubleshooting skills, capable of independently writing, editing, and executing code. It effectively handles code translations, making it especially useful for modernizing legacy applications and migrating existing codebases.
Beyond improvements to the existing model, Anthropic has plans to diversify Claude’s functionalities further. This includes the introduction of new features aimed at broadening its business applications, incorporating integrations with various enterprise systems. A key forthcoming feature is memory functionality, enabling Claude to remember user preferences and interaction histories. This advancement is expected to enhance personalization and efficiency, aligning with similar endeavors by competitors in the space.
As Anthropic looks forward to the potential of Claude, the company is eager to see how users will harness its capabilities for innovative solutions and discoveries.
Ranking
