TIER IV presented five open-source end-to-end driving models for Autoware
The models range from a camera-only design to configurations that add LiDAR, radar and high-definition maps, and include an Autoware integration of NVIDIA's Alpamayo 2 Edge and AutoE2E, which TIER IV leads with the Autoware Foundation.
TIER IV, the Tokyo company behind the Autoware open-source autonomous-driving software, announced on 11 November a set of open-source end-to-end driving models and Autoware integrations, with source code on GitHub and model weights on Hugging Face, according to its press release. Not all were new that day: TIER IV first released METEOR in September, and the Alpamayo 2 Edge weights are NVIDIA’s, published on 10 November. It presented them as outcomes of Autoware 2.0, the architecture that combines AI models with rule-based software.
The announcement covered five models:
- Alpamayo 2 Edge, NVIDIA’s camera-only reasoning model with 2 billion parameters, which TIER IV said it has integrated with Autoware and is now optimising for real vehicles.
- AutoE2E, led by the Autoware Foundation and TIER IV, with 79.9 million parameters. TIER IV said it can run on computing platforms beyond NVIDIA hardware. Only its real-time Reactive Model has been released; the Foundation said it is integrated with Autoware for open-loop evaluation, with closed-loop testing on a real vehicle under way, and a World Model and a Reasoning Model are still in development.
- METEOR, TIER IV’s own 16-million-parameter model aimed mainly at privately owned cars, which uses one to eight cameras without high-definition maps and runs on NVIDIA and Renesas chips, according to the company.
- OnePlanner, which joins LiDAR-based perception with motion planning and uses high-definition maps to plan six seconds ahead.
- LatentDrive, a camera-based world-model approach planning four seconds ahead without high-definition maps, whose integration with Autoware is still under way.
TIER IV said carmakers and component suppliers can adapt the models to their own vehicle specifications. Its chief executive, Shinpei Kato, said carmakers and suppliers need to be able to bring such technology in-house without depending on a particular AI model or semiconductor platform.
Source: TIER IV press release, 11 November 2026
Get the month in vehicle software by email: The SDV Sector Brief ↓