Anthropic's new Claude, Google's coding assistant, and Tencent's 'thinking' AI: This week's AI launches

Anthropic announced Claude 3.7 Sonnet this week, which it called its “most intelligent model to date and the first hybrid reasoning model on the market.” Claude 3.7 Sonnet users through Anthropic’s API can control how long the model “thinks.”

The hybrid reasoning model can either generate a near-instant response, or show its step-by-step thinking process, according to Anthropic.

“Claude 3.7 Sonnet shows particularly strong improvements in coding and front-end web development,” Anthropic said.

The company also introduced Claude Code in a limited research preview — “a command line tool” that allows developers to delegate coding tasks to Claude.

Claude 3.7 Sonnet is available to all tiers of Claude plans, and through Anthropic API, Amazon Bedrock (AMZN), and Google (GOOGL) Cloud’s Vertex AI.

Google made Gemini Code Assist for individuals available in public preview for free this week. Developers around the world can use the coding AI assistant, which is powered by Google’s Gemini 2.0 model. The AI assistant supports all public domain programming languages, and is optimized for coding.

“We’re offering practically unlimited capacity with up to 180,000 code completions per month using Gemini Code Assist — a ceiling so high that even today’s most dedicated professional developers would be hard-pressed to exceed it,” Google said.

Chinese firm Tencent (TCEHY) released its Hunyuan Turbo S AI model this week that it said is a “new generation of fast thinking.” Unlike DeepSeek’s R1 reasoning model and the Hunyuan T1 model “that require ‘thinking before answering,’” the new Turbo S model can “instant reply,” and reduce delay by 44%, the company said on WeChat.

Tencent said Turbo S has demonstrated comparable performance to AI models such as DeepSeek-V3 and OpenAI’s GPT-4o on multiple industry benchmarks in math, reasoning, and other fields.

Turbo S is available through the Tencent Cloud API for developers and enterprise users.

Voice AI startup Hume AI launched Octave TTS this week — a text-to-speech system built on LLM intelligence “that understands what it’s saying.” Octave, which stands for omni-capable text and voice engine, is a speech-language model that can be expressive and nuanced due to its ability to understand words in context, Hume said.

With its AI-powered voice capabilities, Octave can act out characters, generate voices based on prompts, and change the emotion and style of its voice when instructed to by users.

Octave is initially focused on English, but can speak fluent Spanish as well, the company said, adding that it will improve proficiency in other languages.

BigID, a company focused on data security, privacy, compliance, and governance, announced a data security platform called BigID Next this week. The platform is the first cloud-native, AI-powered DSP for enterprises, the company said.

The platform allows enterprises to automate and scale data protection, BigID said. Its features include agentic AI assistants for security and compliance, and automated security and privacy tools.

“With AI reshaping data security and compliance, companies need a solution that’s not just reactive but intelligent, adaptive, and scalable,” Dimitri Sirota, co-founder and CEO of BigID, said in a statement. “BigID Next sets a new standard for how enterprises protect data, reduce risk, and enable innovation—all within a single, unified platform.”

You.com launched its deep research AI agent, ARI, this week, which it said is “the first professional-grade research agent.”

In five minutes, the deep research agent can read and analyze up to 400 sources to generate research reports, the company said. ARI stands for Advanced Research and Insights.

“ARI’s breakthrough is its ability to maintain contextual understanding while processing hundreds of sources simultaneously,” Bryan McCann, co-founder and CTO of You.com, said in a statement. “When combined with chain-of-thought reasoning and extended test-time compute, ARI is able to discover and incorporate adjacent research areas dynamically as analysis progresses.”

AI-powered studying and learning platform, StudyFetch, launched Tutor Me this week. The AI tutor can provide students with real-time, personalized responses in a web conference-style setting.

Tutor Me can quiz students, help find relevant pages in a textbook, and track lesson progress.

“One of our core principles is that every student deserves an opportunity to succeed,” Sam Whitaker, director of social impact at StudyFetch, said in a statement. “Most importantly, they should be able to succeed in the format that best supports their learning. Our mission is to create an equal opportunity for every student, through personalized offerings, affordable prices and innovative technology for all.”