ElevenLabs releases Scribe v3 speech recognition models built for noisy calls
ElevenLabs released Scribe v3 and Scribe v3 Live, speech recognition models designed to handle background noise and crosstalk, with the live version using conversation context so voice agents do not cut callers off. Independent benchmarks show it narrowly beats rivals on accuracy, though some competitors are much faster and about half the price.
Teams building voice agents and transcription tools get a more accurate option, but must weigh it against cheaper, faster alternatives.
aii.news picks and summarizes stories from trusted news sources with the help of AI. Every story links to the original reporting. How we work.
Keep reading
-
Google releases Gemini 4 Hydrogen, now the default model in Search AI Mode
Google launched Gemini 4 Hydrogen, which it says handles multi-part questions, typos and very specific requests better, and it replaces the Flash series. It is available today in Search, the Gemini app, AI Studio, the Gemini API and Gemini for business.
Why it mattersBillions of searches and many developer apps will now run on this model, and independent tests suggest it offers near-top quality for less money than some rivals.
-
Google releases Gemini 4 Carbon, its strongest model yet, with an agent 'swarm' feature
Google is rolling out Gemini 4 Carbon to the general public after first giving it to a select group of cybersecurity defenders. Google says it pushes beyond its Argon model, though Argon remains the suggested choice for most coding and knowledge work.
Why it mattersDevelopers and users now have a new top-tier Google model to compare against the best systems from Anthropic and OpenAI.
-
Anthropic releases Claude Fable 5.2 and Mythos 5.2 and cuts API prices by 20%
Anthropic introduced Claude Fable 5.2 and Mythos 5.2, which share one underlying model with different safeguards and bring gains in coding, reasoning and vision. Fable is broadly available, while Mythos is limited to vetted cybersecurity and life sciences groups, and both now cost $8 per million input tokens and $40 per million output tokens.
Why it mattersDevelopers get a stronger coding model at a lower price, and the split release shows how labs are gating the most capable models for sensitive uses.
-
OpenAI research explains why frontier models' reasoning is becoming harder to monitor
OpenAI's safety team studied why chains of thought in models like GPT-6 Astra increasingly evade oversight and found four training steps, including architecture choices and reinforcement learning, drive the loss. The analysis says this reflects tradeoffs around capability and compute cost rather than an unavoidable trend.
Why it mattersReading a model's reasoning is a key safety check, and this shows labs can choose to preserve it if they accept some cost.
-
OpenAI launches Daybreak security tools that use AI agents to pentest apps and prepare fixes
OpenAI introduced Daybreak Red Team, which uses coordinated AI agents to test web applications for weaknesses and is offered through early access. It also launched Defense Factory, which investigates security findings and prepares patches, linking vulnerability discovery with fixing and retesting.
Why it mattersSecurity teams may soon be able to hand both finding and patching of app flaws to AI agents, which could speed up fixes but needs careful human review.
