Anthropic launches new Claude AI models — beating ChatGPT in some tasks

Anthropic launches new Claude AI models — beating ChatGPT in some tasks

Anthropic — one of the leading AI-companies, has announced the release of its new Claude 3 model family, setting new industry benchmarks across a wide range of cognitive tasks. The family includes three state-of-the-art models: Claude 3 Haiku, Claude 3 Sonnet, and Claude 3 Opus, each offering increasing levels of capability and performance.

The Claude 3 models have demonstrated near-human levels of comprehension and fluency on complex tasks, outperforming ChatGPT and Gemini on common evaluation benchmarks such as undergraduate level expert knowledge (MMLU), graduate level expert reasoning (GPQA), and basic mathematics (GSM8K). The models excel in analysis, forecasting, content creation, code generation, and multi-language conversation.

In addition to their impressive intelligence, the Claude 3 models boast near-instant response times. Haiku, the fastest and most cost-effective model in its category, can read and process a 10,000-token research paper with charts and graphs in under three seconds. Sonnet delivers speeds twice as fast as its predecessors, while Opus maintains similar speeds to Claude 2 and 2.1 but with significantly higher intelligence levels.

The new models also showcase strong vision capabilities, processing a wide range of visual formats including photos, charts, graphs, and technical diagrams. This feature is particularly valuable for enterprise customers with knowledge bases encoded in various formats.

Anthropic has addressed previous issues with unnecessary refusals, improving the models' contextual understanding and reducing instances of harmless prompt refusals. The Claude 3 models also demonstrate a twofold improvement in accuracy on complex, factual questions compared to Claude 2.1, while reducing incorrect answers.

With a focus on responsible design, the Claude 3 models have been developed to be trustworthy and capable. Anthropic aims to mitigate risks such as misinformation, CSAM, biological misuse, and election interference. The models have also been tuned to address privacy concerns and show less bias than previous iterations.

Claude 3 Opus and Sonnet are now available through the Claude API, which is generally available in 159 countries. Sonnet powers the free experience on, while Opus is available for Claude Pro subscribers. Haiku will be available soon.

As Anthropic continues to push the boundaries of AI capabilities, they remain committed to ensuring that safety guardrails keep pace with these advancements. The company believes that being at the forefront of AI development is the most effective way to steer its trajectory towards positive societal outcomes.