Nvidia open-sources Flora for robot video labeling at 125x lower cost
In brief
- Flora labels one hour of robot video for 72 cents instead of $88 with GPT-6 Astra.
- The system uses three fine-tuned Qwen3.5 models and outperforms competitors in detecting action errors.
- Nvidia positions Flora as a key tool to control the robotics value chain.
Nvidia introduces Flora, an open-source data engine that labels one hour of robot video for 72 cents instead of $88 with GPT-6 Astra, while maintaining slightly higher quality. The system relies on three fine-tuned Qwen3.5 models and detects robot action errors more reliably than competitors, though it may lack descriptive precision according to the announcement.
What it changes for you
- Industry & ManufacturingFlora drastically cuts the cost of video labeling for robots, speeding up the training of industrial vision models.
This summary gives you the essentials. The full article, its figures and quotes are on The Decoder.
Read the original articleAI-generated summary of the source, checked by rules. How we write · Report an error