FrontierAugust 5, 2026via MarkTechPost
NVIDIA Releases Alpamayo 2 Super: A 34B Open Vision-Language-Action Model for Robotaxis and Autonomous Driving Under OpenMDW-1.1
Why it matters
A major open-weight release combining vision, language, and action prediction in a single model under a commercial-friendly license signals NVIDIA's push to democratize the robotaxis stack and compete with closed proprietary systems.
Key signals
- Model: Alpamayo 2 Super, 34B parameters (32B Cosmos 3 Super Reasoner + 2.3B diffusion action decoder)
- License: OpenMDW-1.1 (permissive; allows fine-tuning, derivatives, commercial redistribution)
- Benchmark: 79.2 on LingoQA
- Capabilities: trajectory generation, Chain-of-Causation traces, meta-actions, auto-labels, grounded VQA in single inference pass
- Domain: autonomous driving and robotaxis
- Release date: August 5, 2026
The hook
NVIDIA open-sources a 34B vision-language-action model for autonomous driving—permissive license, multimodal reasoning in one pass.
NVIDIA released Alpamayo 2 Super, a 34B vision-language-action model for autonomous driving, under OpenMDW-1.1 — a permissive license covering fine-tuning, derivatives and commercial redistribution. It pairs a 32B Cosmos 3 Super Reasoner backbone with a 2.3B diffusion action decoder, scores 79.2 on …