Launch report · provisional · permalink
+1 day · 2026-09-23+7 days · 2026-09-29+30 days · 2026-10-22then the normal re-check cycle
What the lab claims, and what independent runs found
Hollow marks are the lab's own figures from its launch post; filled marks are independent runs. A missing mark is missing evidence, never a guess. Benchmark names are fictional.
Where it stands, per domain, against its closest competitors
Each bar is the capability estimate's interval; the dot is the estimate. Overlapping intervals mean the order is not settled yet. Evidence per domain is shown unblended, as in slice 1.
Trade-offs: capability against cost per task
Open this chart in the trade-off canvas and apply your own constraints →
What the lab didn't report
Still unknown
Social posts are views of this report
Three cards at each network's recommended size, drawn by the same chart components as the page above. Each carries its spec: the link opens the downselect prototypes with that exact spec, so a reader can add their own constraints. Nothing is posted before launch.
Card theme