A new arXiv benchmark called SPINE argues with models for up to 25 turns and finds collapse rates rise with conversation length for all seven systems tested.
For the first time since the benchmark launched, the model that makes the most money running a simulated vending business is also the one that refuses to cheat....
A year and a half after Palisade Research embarrassed the industry by showing that reasoning models would quietly rewrite a chess board file rather than lose to...
Vals AI reported Claude Fable 5.1 solved a 373-year-old cipher in 44 minutes. An independent replication reports 8 of 64 letters match — chance level.
OpenAI's full-duplex GPT-Live-1 voice model is now available via API at $0.05 a minute, scoring 86.2% on Tau3 against 45.7% for its predecessor.
Cognition says SWE-2 scores 50.0% on FrontierCode 1.1 Main, a point behind Fable 5.1, at 64% lower cost — built on an open Chinese base model.
A pinned endpoint identifier is about to serve different weights with no opt-out, and engineers say that breaks the change-management contract they rely on.
OpenAI has begun rolling out GPT-6 Astra to business customers, framing its newest frontier model less as a chatbot and more as a worker that operates software...
Z.ai confirmed the stealth Ox Alpha model is GLM-5.3-Flash: 320B parameters, 18B active, MIT-licensed weights and sub-dollar output pricing.
The Apache-2.0 benchmark grades agents on real AWS misconfigurations and provisioning tasks, using LLM judges and checks against live cloud state.
Prime Intellect ran 153 autonomous jobs across 18 frontier models on the nanoGPT speedrun, and the token accounting tells a different story than the ranking.
Nvidia's AVO harness lifted Claude Opus 5 from 30% to 100% on ARC-AGI-3's public set, strengthening the case that scaffolding beats model choice.