← Leaderboard

FLUX-mimic

World-action
mimic · Jul 23, 2026

Video-action model from mimic built on Black Forest Labs’ FLUX 3 video backbone, for dexterous industrial manipulation. Results are self-reported on mimic’s own factory tasks (paper Table I: 82.8% average on the dexterous Franka setup vs 27.6% for a fine-tuned π0.5), not on a public benchmark, so it has no ranked scores yet. Announced as a preview; no weights released.

Type
World-action
Parameters
—
Backbone
FLUX 3 video backbone + action expert
Weights
Unknown
License
–
Commercial use
Unclear
Released
Jul 23, 2026
Robots
Franka (bimanual, dexterous hands), Trossen WidowX (bimanual)
SO-101
–
LeRobot
–

Benchmark results

0 of 13 benchmarks

No benchmark results recorded yet.

Not reported on 13 benchmarks

Running the model

HardwarePrecisionConfigLatency / chunkStepsVRAMChunkControlSource
RTX 5090–preview; one backbone pass per chunk, quantized video backbone, real-time chunking–––2015 HzNov 2026 ↗

Runs on

Wrong or missing number? Report an error