Summary
According to a post on X by @amanvaibhav, Runway released Praxis-1, a 27-billion-parameter world action model, under the Apache 2.0 license. The thread says it tops NVIDIA's RoboLab-120 benchmark at 46.2%. It also reports that, when fine-tuned on Google's RT-1 robot, a platform the model never trained on, it reached 80% on SimplerEnv versus 52.4% for the original policy. The snippet cuts off while describing a claim that web video can train a policy.
AI-assisted summary based on the listed source.
What happened
According to a thread by Aman Vaibhav, Runway released Praxis-1, a 27-billion-parameter world action model, under the Apache 2.0 license, and it tops NVIDIA's RoboLab-120 benchmark at 46.2%. The thread says that, fine-tuned on Google's RT-1 robot, which it never trained on, it reached 80% on SimplerEnv versus...
Why it matters
Robotics researchers and developers could get a large open-licensed model to test and build on, if the release is confirmed. The benchmark and transfer results come from a single thread, so independent evaluations will be important. Watch for the official Runway release, model weights, and third-party results on RoboLab-120 and SimplerEnv.
Signal Intelligence
Signal Strength 91%
Technical label WATCH
Public Interest 40
Category OPEN SOURCE
Reader Depth TECHNICAL
Signal Strength reflects source quality, relevance, freshness and evidence. Public Interest helps organize discovery; it is not proof of truth.
Public Interest components
Recognizable Entity Score 0
Practical Impact Score 26
Novelty Interest Score 94
Consequence Score 50
Curiosity Score 32
Shareability Score 50
Why this is here
VQV surfaced this signal because it is recent, relevant to Open Source LLMs, connected to Scobleizer News Pulse (X).