Anthropic, a leading artificial intelligence company in the field of LLM, has unveiled the Claude 3 model. The key highlight of this flagship model is its exceptional performance, outscoring GPT-4 in almost every test set and also surpassing Gemini Ultra.
Claude 3 consists of three sub-models, namely Opus, which outperformed GPT-4, Sonnet, closely resembling GPT-4, and Haiku, positioned above GPT-3.5. The Haiku model is designed to run tasks swiftly; for instance, processing documents exceeding ten thousand tokens takes less than 3 seconds. All three models support reading images, document files, and various flowcharts, achieving scores close to Gemini Ultra.
Each sub-model of Claude 3 supports a context window of 200,000 tokens, but there is a 1M model available for individual usage. The development in this version enables precise information retrieval in every text point. In the case of the Needle In A Haystack test, where text is deliberately concealed, Claude 3 can accurately identify anomalies where text has been misplaced.
The pricing for Claude 3 is as follows:
Opus: $15 per 1M token input and $75 per 1M token output
Sonnet: $3 per 1M token input and $15 per 1M token output
Haiku: $0.25 per 1M token input and $1.25 per 1M token output
Currently, Anthropic offers Opus and Sonnet exclusively, while Amazon Bedrock provides only Sonnet and Google Cloud offers limited access to Sonnet.
In summary, Anthropic’s Claude 3 is a cutting-edge AI model that sets a new standard in artificial intelligence capabilities.
TLDR: Anthropic introduces Claude 3, a powerful AI model outperforming GPT-4 and Gemini Ultra, with impressive speed and accuracy in various tasks. Pricing varies for Opus, Sonnet, and Haiku, and different service providers offer access to specific sub-models.
Leave a Comment