TIER IV Releases Five E2E Autonomous Driving AI Models With Selectable Map and Sensor Configurations

prtimes.jp reported on November 11, 2026, that TIER IV had released five E2E autonomous driving AI models with selectable map and sensor configurations. We have not been able to verify the primary source, so please refer to the original for details.

TIER IV released models ranging from one that uses only cameras to models that also use LiDAR and radar. For METEOR, it said the model supports different semiconductor environments, including those from NVIDIA and Renesas. Automakers and parts suppliers can consider the models’ input requirements and semiconductor environments in light of their vehicle specifications, but will still need to separately assess performance and safety in automotive environments before deciding whether to implement them.

The new aspect of this announcement is that TIER IV released a group of models with different sensor inputs, levels of map dependence and supported semiconductor environments, rather than presenting a single E2E model. Among the models with published figures, sizes range from 16 million parameters for METEOR to 79.9 million for AutoE2E and 2 billion for Alpamayo 2 Edge. The lineup spans lightweight camera-only configurations through configurations that use multiple sensors or reasoning, suggesting that models can be selected according to the application and automotive environment. At the same time, Autoware 2.0 aims to provide an extensible platform combining AI models with conventional rule-based technology, extending efforts to build on existing software platforms. An article we published on Nov. 10, NVIDIA Explains In-Vehicle Compute Resources and Inference Models for Autonomous Driving Under Latency Constraints, addressed in-vehicle compute resources and latency constraints. This announcement outlines options for model size and input modes, but does not by itself establish processing speed or latency within in-vehicle compute limits. How differences in model size align with actual vehicle requirements must be considered alongside runtime compute load and evaluation results. Structurally, the announcement positions Autoware as a point of contact for releasing and integrating models, with automakers and others optimizing them to meet vehicle specifications and performance requirements. METEOR cites support for different semiconductor environments, including both NVIDIA and Renesas, offering options that are not tied to a particular computing platform. However, an open-source model is not necessarily ready to be implemented as-is in each company’s vehicles. Alpamayo 2 Edge is under development for vehicle adaptation, while LatentDrive is being tested in automotive environments; the models are also at different stages of development. The next points to check are each model’s compute requirements and latency during in-vehicle operation, the conditions used to assess driving performance and safety, and the specific specifications of supported automotive computers. The announcement did not provide comparable quantitative figures on these points. AutoE2E development is continuing beyond the Reactive Model, so the release and integration status of the World Model and Reasoning Model will also be a focus in distinguishing planned capabilities from those currently available.

METEOR’s stated support for different semiconductor environments, including Renesas, indicates that autonomous driving systems using Japanese automotive semiconductors could be optimized based on publicly released models. However, the announcement did not specify operating performance on Renesas products or any adoption track record.