What was released
Runway has released its robot model Praxis-1 with open model weights and source code. According to Runway, both are available under the Apache 2.0 license, which allows commercial use. The company first introduced Praxis-1 in late September but initially gave access only to select partners. With the release, Runway publishes a detailed research report with new performance tests.
The model takes in camera images, text instructions, and information about a robot's position. From those inputs, it predicts how the environment will change and calculates matching robot movements. New camera frames continuously feed into the control.
Video training as the basis
Praxis-1 builds on a video model that learned from ordinary footage how objects move and interact. Runway then fine-tuned it with videos of human actions and robot data. The company calls the model a World Action Model. Whether this approach actually reduces the need for robot data that has to be collected specifically remains open: to adapt it to a Franka robot arm, Runway used about 277 hours of recordings from the public DROID dataset, according to its research report, and 1,400 to 2,600 hours for a dual-arm robot.
Performance figures from the developer
On Nvidia's RoboLab-120 leaderboard, Praxis-1 scores a 46.2 percent success rate; FLUX 3 Action, also built as a World Action Model, scores 42.9 percent. The benchmark covers 120 tasks in simulation. On a real Franka arm, across ten tasks with ten attempts each, Praxis-1 succeeded 88 times versus 80 for FLUX 3 Action. For a previously unseen mobile robot, Runway reports an average success rate of 80 percent in simulated tests, while Google's RT-1 scores 52.4 percent on the same benchmark, according to Runway. All results were produced by Runway and have not yet been independently verified.
Limits
Runway names weaknesses itself: the model occasionally confuses similar-looking objects, struggles with vague instructions, and cannot yet reliably plan complex action sequences on its own. Longer tasks should therefore be broken into explicit steps. The base model also needs additional training for new robot systems. According to Runway, the standard version needs a GPU with 80 GB of memory; a compressed version should fit on an RTX 5090 with 32 GB, but the actual peak memory usage has not yet been measured according to the GitHub documentation.
Assessment
For companies working on robotics, an openly licensed base model could lower the entry barrier, provided the performance claims are independently confirmed. Whether the data requirement really drops is not established by the current sources. Teams that want to run it locally or on their own hardware should check the stated memory requirements in advance.
Sources
Editorially owned by Ideal Syka. Sources and method: Newsroom & method. Tips and corrections: ai@i6eal.de.
