Tags
"Prompt custom vocal personas, direct line-by-line delivery, and add vocal bursts," Google says of Gemini 3.8 Flash TTS, which is now live in Google AI Studio and the APIs alongside Gemini 3.8 Flash-Lite TTS.
"Passing the savings directly to you," OpenAI says of GPT-6 Sol at $2/$10 per million tokens, half the price of GPT-5.6 Sol, and this time the lower price isn't a promo.
An early tester pointed Claude Opus 5.5 at a 680,000-line code migration and had it done in under a day. It also runs 40% cheaper than Opus 5 at default settings, with cache reads down 60% to $0.20 per million.
Anyone pulling Xiaomi's fresh MiMo-V2.6-Pro weights today also gets a claim: number 6 on the Artificial Analysis Intelligence Index. The ranking traces back to a post dated September 13, 2025.
"Close to Opus 5 performance on certain tasks at a price of Grok 4.6" is how the SpaceXAI post sums up Grok 4.7, now on Grok Build, Cursor and the APIs, though no benchmark numbers travel with the claim.
Transcribe a call and the text marks who said what, with the ums stripped out. That's xAI's Grok Voice Transcribe 2.0, first for accuracy among 32 streaming models on the Artificial Analysis leaderboard.
Alibaba shipped Qwen3.8-Omni-Flash, which reads video and audio alongside text and images inside a 1M-token window. Alibaba claims performance close to Gemini 3.8 Flash on two multimodal benchmarks and publishes no numbers.
Jev replies with a typed decision and a score "Why have superhuman chat models not led to AGI?" Diogo Almeida asked that, then spent two years in stealth on Jev, a model that hands back one option from a developer-defined list instead of text.
Gemini 3.8 Live Extended Thinking (High) claims first place on the Speech to Speech Index, ahead of GPT Live 1, which OpenAI opened to API developers on September 10. Google is rolling it out on APIs, AI Studio and Gemini Live.
"Machine translation is still broken for most of the world's languages," Cohere co-founder Nick Frosst told The New Stack. His answer, North Small Translate, covers 50 languages, with weights free for noncommercial use only.
"The smallest model in DeepSeek new architecture family" runs 552B parameters as a mixture of experts on a new encoder-decoder structure. V4.1 Flash weights are on Hugging Face after a beta window set to close September 10.
OpenAI's Critical cyber threshold means finding and building functional zero-day exploits "without human intervention", and BleepingComputer reports GPT-6 Astra clears it while being harder to monitor.
Inception shipped Mercury 2.5, a diffusion model it clocks at 1,107 tok/s on widely available NVIDIA GPUs, and says it matches Gemini 3.5 Flash and GPT 5.6 Luna Low on quality, a comparison resting on its own benchmarks.
Sketch a rough layout, hand it to ChatGPT Images 2.5 and mark where the change goes. OpenAI says the new model renders up to 50% faster, and Axios found it held the likeness of people and pets better in early testing.
IFM published the training stages where its own models gamed evaluations, copying hidden answers and exploiting checkers, and shipped those checkpoints with K2 Horizon, six models from 0.9B to 375B under Apache 2.0.
Microsoft cut its speech-to-text price from $0.36 an hour to $0.10, roughly 72% in five months, and bundled speaker diarization, timestamps and 60 languages into that launch rate.
The API got GPT-6 Astra before Codex did. Access in ChatGPT Work and Codex may take a few days to reach Plus and all Business users, at $10 per million input tokens and $50 per million output.
"Enhanced audio fidelity and vocal clarity" is Google's pitch for Lyria 3.5, the full-song music model now live in AI Studio, the Gemini app and the APIs after its July debut in Flow Music.
"Welcome to the AGI era," OpenAI President Greg Brockman said as GPT-6 Astra launched, though access starts with enterprise customers in the Daybreak program, with paid tiers, the API and AWS following in the coming days.
Developers hitting Meta's model APIs today get Muse Spark 1.3, which Meta says outscored GPT-5.6 and Opus 5 on DeepSWE 1.1. An open-weight version and an upgrade codenamed "Watermelon" are next in line.
