DrivingBench evaluates whether frontier AI models like GPT-6 Astra can drive a real car by controlling a Toyota Corolla on a predefined course, tracking metrics such as progress, distance, and token cost.
# DrivingBench
Source: [https://drivingbench.com/](https://drivingbench.com/)
Can frontier models drive a*real car*? We give them control of a Toyota Corolla’s steering, accelerator, and brakes, then evaluate them on a fixed cone course\.
[Follow @drivingbench](https://x.com/drivingbench)[Report](https://drivingbench.com/report/)[Eval traces](https://drivingbench.com/downloads/)
## Leaderboard
attemptprogressHow far along the course centerline the attempt got while staying within 4 m of it, as a share of the centerline length to the finish zone; a collision keeps the progress reached before it\.distanceGPS speed integrated from the first accepted set\_motion to the end of the attempt's last engagement\.finish timeFirst accepted set\_motion to the end of the attempt's last engagement\. Shown for completed runs only; DNF means did not finish\.commandsAccepted set\_motion and stop\_now calls during the attempt\.tokens · costTotal tokens and cost at list prices for the attempt, including the reflection after it\.
attemptprogressHow far along the course centerline the attempt got while staying within 4 m of it, as a share of the centerline length to the finish zone; a collision keeps the progress reached before it\.distanceGPS speed integrated from the first accepted set\_motion to the end of the attempt's last engagement\.finish timeFirst accepted set\_motion to the end of the attempt's last engagement\. Shown for completed runs only; DNF means did not finish\.commandsAccepted set\_motion and stop\_now calls during the attempt\.tokens · costTotal tokens and cost at list prices for the attempt, including the reflection after it\.
attemptprogressHow far along the course centerline the attempt got while staying within 4 m of it, as a share of the centerline length to the finish zone; a collision keeps the progress reached before it\.distanceGPS speed integrated from the first accepted set\_motion to the end of the attempt's last engagement\.finish timeFirst accepted set\_motion to the end of the attempt's last engagement\. Shown for completed runs only; DNF means did not finish\.commandsAccepted set\_motion and stop\_now calls during the attempt\.tokens · costTotal tokens and cost at list prices for the attempt, including the reflection after it\.
attemptprogressHow far along the course centerline the attempt got while staying within 4 m of it, as a share of the centerline length to the finish zone; a collision keeps the progress reached before it\.distanceGPS speed integrated from the first accepted set\_motion to the end of the attempt's last engagement\.finish timeFirst accepted set\_motion to the end of the attempt's last engagement\. Shown for completed runs only; DNF means did not finish\.commandsAccepted set\_motion and stop\_now calls during the attempt\.tokens · costTotal tokens and cost at list prices for the attempt, including the reflection after it\.
[**Explore the track**View the trajectory replays](https://drivingbench.com/#course)## Explore the course
OpenAI releases GPT-6 Astra, its most capable AI model for advanced reasoning, software engineering, and agent workflows, featuring async tool calling and mid-turn steering with a tiered rollout plan.
OpenAI introduces GPT-6 Astra, a state-of-the-art AI model that achieves top scores on multiple benchmarks and excels in computer use, alignment, and professional tasks, with rollout to users starting soon.
GPT-6-Astra is a major AI model release from OpenAI with significant advancements in capabilities, particularly for ambitious projects, 3D tasks, and computer use, outperforming previous models in various benchmarks.
GPT-6 Astra has successfully beaten all 48 levels of the 'I'm Not A Robot' game, demonstrating advanced AI capabilities in interactive challenge environments.