ChatGPT gets more natural voice, Grok launches powerful 4.5 model

ChatGPT upgraded its voice mode, Grok launched a powerful new model, and Perplexity added new search controls.

ChatGPT

1 update

Expanded summary · ChatGPT

You can now have more natural, live conversations with ChatGPT thanks to new voice models.

OpenAI today released new conversational models, called GPT-Live-1 and GPT-Live-1 mini, claiming that they sound more natural and can handle turn-taking better. The company is also replacing its current Advanced Voice Mode in ChatGPT with GPT-Live-1 mini by default.

Users of paid tiers will be able to access the larger GPT-Live-1 model…

The app's Advanced Voice Mode now uses a smaller, faster model by default.

Original reportingOpenAI releases new voice models for more natural live conversationsRead original ↗

Perplexity

1 update

Expanded summary · Perplexity

You can now set a limit on how long your search results are, which helps you manage costs.

Smart media selection: Intelligently chooses between images, videos, or both based on query type Configurable control: Enable/disable and override media types as needed New max_tokens parameter: Control the maximum tokens extracted per page in search results. This gives you finer control over response size and costs.

Learn more →

Original reportingChangelogRead original ↗

Grok

1 update

Expanded summary · Grok

Grok has released its newest model, Grok 4.5, which is designed to be very powerful.

SpaceXAI has released its latest model, Grok 4.5 — the first since the company went public several weeks ago.

The company released benchmark metrics Wednesday that appeared to show Grok’s competitiveness with other top models from SpaceXAI competitors, although just short of best-in-class: This new version can handle many different tasks, like helping with coding, office work, research, and writing.

Original reportingSpaceXAI releases Grok 4.5, which Elon describes as an 'Opus-class model'Read original ↗

Claude Code

1 update

Expanded summary · Claude Code

The inside story of how Claude Code went from an internal CLI to Anthropic's coding agent, told by researchers, engineers and early users who built it.

The Making of Claude Code

Original reportingThe Making of Claude CodeRead original ↗

Claude

3 updates

Expanded summary · Claude

UPDATE: Golden Gate Claude was online for a 24-hour period as a research demo and is no longer available. If you'd like to

UPDATE: Golden Gate Claude was online for a 24-hour period as a research demo and is no longer available. If you'd like to find out more about our research on interpretability and the activation of features within Claude, please see t…

On Tuesday, we released a major new research paper on interpreting large language models, in which we began to map out the inner workings of our AI model, Claude 3 Sonnet.

Original reportingGolden Gate ClaudeRead original ↗

Expanded summary · Claude

In this post, we are sharing what we have learned about the trajectory of potential national security risks from frontier AI models, along with some of our thoughts about challenges and best practices in evaluating these risks.

The informa…

While AI capabilities are advancing quickly in many areas, it's important to note that real-world risks depend on multiple factors beyond AI itself.

Original reportingProgress from our Frontier Red TeamRead original ↗

Expanded summary · Claude

At the end of the project, we now have more experiments and evaluations we’d like to run than we started with.

For example, we think a very important experiment to repeatedly run will be to measure the speedup that LLMs might provide towar…

This empirical work confirms that frontier threats red teaming in areas of national security is important and timely.

Original reportingFrontier threats red teaming for AI safetyRead original ↗