TLDRocket
Sign in

Evaluators stress-tested OpenAI’s GPT-6 Astra on one-shot build tasks and a physical driving benchmark

Benchmark result Provisional 78% confidence first seen

Two outlets reported results from stress tests of GPT-6 Astra, including performing multiple one-shot build tasks with minimal steering. One test involved the model completing a Toyota Corolla obstacle course in the DrivingBench benchmark, after which coverage emphasized the model’s ability to control physical systems in addition to software-style tasks.

Claude Fable 5.1 Corey GPT-6 Astra Grant OpenAI Anthropic Blender Cat Doom DrivingBench RoboHarm Robocurve Toyota Corolla xAI

Source coverage

GPT-6 Astra and Claude Fable Rarely Refuse Dangerous Robot Commands, Test Finds The Decoder newsletter GPT-6 Astra stress test: six one-shot build tasks YouTube newsletter OpenAI’s GPT-6 Astra Just Drove a Real Car Through an Obstacle Course Trending Topics independent

The daily briefing

Every AI story that matters, in your inbox by 8am.

TLDRocket reads all relevant sources, removes duplicate coverage, and summarises the day in two minutes. Follow companies and topics for alerts, or get the briefing in Slack. Free, no spam, unsubscribe anytime.