OpenAI's Astra model went for a drive and no one died
First LLM to pass test did it in a parking lot at snail's pace, at considerable expense, but it still counts for something
OpenAI's GPT-6 Astra model successfully drove a car around a set of cones in a parking lot, marking a significant milestone in the field of generative AI. The model completed the course in just 5 minutes and 22 seconds, despite being one of the most expensive attempts to date. Researchers Tobias Gessler, Aditya Ramabadran, and Simon Mahns spent $7.74 to generate 6.6 million tokens for inference, costing ~$92.47 per mile.
Despite the high cost, the researchers found that specialized driving models are still preferred for real-world applications, with GPT-6 Astra needing datacenter GPUs for inference. However, the study suggests that as technology improves, general-purpose models like GPT-6 Astra may become more capable and potentially fulfill driving tasks in the future.
Written by urgent.news from The Register's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
Also reported by 2 other outlets
- OpenAI's Astra model went for a drive and no one died theregister.com
- Airbnb expands deal with OpenAI for GPT-6 Astra access techinasia.com
