Tsinghua · Galaxea

Fast-WAM↗

Fast-WAM is a WAM from Tsinghua · Galaxea, released March 17, 2026. It has no overall rank on the Robot Intelligence Index, which needs both a simulation and a real-robot result. Its best result is #7 of 20 on LIBERO-Pro2. World-action model that uses a video DiT as a single-pass encoder, skipping future-frame generation at test time.

Release date: Mar 17, 2026

Type
WAM
Model weights
Open
License
—
Parameters
6B
Pricing
Self-hosted
Benchmarks
4
Overall rank
Not ranked
Needs a real-robot result to rank overall.
Real robots rank
Not ranked
No real-robot results yet.
Simulation rank
#36of 49
Rating 1302 · range 21–38
Rating
1332± 25.4
10601989
Cost / task · est.
$0.028
$0.0087$17.74
Time / task · est.
30s
9.5 s1 h 22 min

Per task attempt; the tiles show the median across benchmarks. Self-hosted compute is priced at one H100 at $3.29 per hour (Lambda on-demand, Oct 2026).

Inference

Paper/project: 190 ms per inference on one RTX 5090D V2 (IDM variant 810 ms). Repo update: compiled path 110 ms on RTX 4090, 210 ms on H20. Size = Wan2.2-5B video DiT + 1B action expert (sum of components).