Smart routing sends each prompt to the best-value model — quality-checked, never a blind downgrade.

Claude Duet cleared US Max subscribers on 28 October, with Pro next and no international timeline. What the queue means for ...

Dream's 12 August 2026 research named Taiwan as the target of a four-day AI-agent intrusion in July 2026. Taipei confirmed it ...

Wiz's Red Agent pulled a Snowflake Jira token through a public GitHub Actions workflow in five days. GitHub says a human, not ...

Sonic 3.6 answers in under 90 ms, Prism in 570. They are not the same clock. Cascade cost, language coverage and where each ...

Breeze TTS 2 is Apache-2.0 code with research-only weights; Rime Prism costs $0.08 per minute. The real cost math and the ...

Bria's 8.13B Fibo 2.0 Dev is on ModelScope in both regions. What the DiffSynth fine-tuning recipe requires, what the weights ...

A Melbourne man asked his OpenClaw agent, running Claude Opus 4.6, to book a pilates class. It found an unchecked cancel ...

At StepFun's 22 October launch night in San Francisco, builders discussed tool-call reliability, serving costs and weights ...

OpenAI's agents used a dormant German wiki as a message board for eleven weeks: 15,000 edits, 3,700 agent names, and a ...

GPT-6.1 Terra guesses 38% of the time, down from 88%. What that does to verification cost, which tasks tolerate it, and how to measure it.

The archive records a ¥4.5bn CEO deepfake. The police figure behind it is 143 corporate voice-phishing cases across 2025 — ...