Wednesday, 26 November 2025
29.3 C
Singapore
23.4 C
Thailand
27.6 C
Indonesia
28 C
Philippines

Anthropic’s newest Claude chatbot outperforms GPT-4o in benchmarks

Explore Claude 3.5 Sonnet, Anthropic's latest AI model, now available. It excels at understanding nuance and visual input, outpacing GPT-4o benchmarks.

On Thursday, Anthropic introduced its latest AI language model, Claude 3.5 Sonnet. This new version surpasses the company’s previous top-tier model, the Claude 3 Opus, while operating at twice the speed. You can now explore this enhanced chatbot, even with a free account.

Key features and performance

Claude 3.5 Sonnet is the first in the Claude 3.5 series and is considered Anthropic’s most balanced model. Future releases in this series will include Claude 3.5 Haiku, the fastest model, and Claude 3.5 Opus, the most powerful. These updates will roll out later this year while the current versions remain on Claude 3. The quick release of Sonnet, just months after the Claude 3 family, highlights the rapid pace at which AI companies are developing their technologies.

Anthropic claims that Claude 3.5 Sonnet significantly improves understanding of nuance, humour, and complex prompts, enabling it to write in a more natural tone. Benchmark tests indicate that the new model sets industry records for graduate-level reasoning, undergraduate-level knowledge, and coding proficiency. It surpasses OpenAI’s GPT-4o in many of these benchmarks. However, it is worth noting that the latest models from Claude, ChatGPT, Gemini, and Llama are all closely matched, often scoring within a few percentage points of each other, reflecting the intense competition in the AI field.

Enhanced visual interpretation and a new workspace

The company asserts that Claude 3.5 Sonnet excels at interpreting visual input better than its predecessor, Claude 3.0 Opus. The new model can accurately transcribe text from imperfect images, a feature expected to attract retail, logistics, and financial services customers who require precise data interpretation from charts, graphs, and other visual cues.

Claude’s latest update also includes a new workspace feature called Artifacts. When you prompt the chatbot to generate content such as code, text documents, or web designs, a dedicated window appears next to the chat interface. This Artefacts window allows you to request changes, and it will update with the chatbot’s latest output. Anthropic sees artefacts as a step towards making Claude a hub for broader team collaboration. The company envisions a future where teams and entire organisations can securely centralise their knowledge, documents, and ongoing projects in one shared space, with Claude acting as an on-demand team member.

Availability and pricing

Claude 3.5 Sonnet is now available for anyone with an account to try on Anthropic’s website and through the Claude iOS app. Pro and team subscribers on these platforms will benefit from higher token counts. Additionally, you can access it via the Anthropic API, Amazon Bedrock, and Google Cloud’s Vertex AI. The cost remains the same as the previous model, at US$3 per million input tokens and US$15 per million output tokens.

Hot this week

Heidi launches in Singapore after securing US$65 million in Series B funding

Heidi opens its Singapore hub after raising US$65 million, aiming to expand healthcare AI adoption across Southeast Asia.

Roblox’s selfie verification hints at a more intrusive online future

Roblox’s new age verification system signals a growing shift toward identity checks across online platforms, raising safety and privacy concerns.

OpenAI was blocked from using the term ‘cameo’ in Sora after a temporary court order

A judge blocks OpenAI from using the term “cameo” in Sora until 22 December as Cameo pursues its trademark dispute.

TikTok tests new tools to help users manage AI-generated content

TikTok tests an AI content slider and invisible watermarks to help users control and identify AI-generated videos on the platform.

Meta explores an AI briefing tool aimed at Facebook users

Meta is developing Project Luna, an AI tool designed to deliver personalised morning Facebook briefings to users.

DBCS launches global design platform and unveils SG Mark 2025 winners

DBCS celebrates 40 years with the launch of WDBO and SG Mark 2025, spotlighting Singapore’s role in global design and innovation.

Chrome tests new privacy feature to limit precise location sharing on Android

Chrome for Android tests a new privacy feature that lets websites access only approximate location data instead of precise GPS information.

OpenAI introduces a new shopping assistant in ChatGPT

OpenAI launches a new ChatGPT shopping assistant that helps users compare products, find deals, and search for images ahead of Black Friday.

OpenAI was blocked from using the term ‘cameo’ in Sora after a temporary court order

A judge blocks OpenAI from using the term “cameo” in Sora until 22 December as Cameo pursues its trademark dispute.

Related Articles