NEW RUNTIME
Raw SignalOct 10, 20261 min read1 source

Checkpoints available in both JAX and PyTorch via Hugging Face .

In one minute

Google’s Hugging Face model card lists official pretrained checkpoints for the Tracking Any Point (TAP) family from Google DeepMind and Google Research.

The family tracks a queried point on a physical surface across video frames and estimates visibility. The catalog includes TAPNext++, TAPIR, BootsTAPIR, and other models. Listed checkpoints span JAX and PyTorch; the page also documents online and offline variants for some models. The card states that the checkpoints are licensed under Apache 2.0. This is an implementation and distribution update for developers evaluating point tracking, not evidence of a new model release or a comparative performance advance. Source: https://huggingface.co/google/tapnet

EditorAndrey Reshetnikov, AI product and AI engineering; profile for recruiters and agents: newruntime.com/owner-profile.json

  1. Oct 16, 2026Salience must prove its switch improves the whole fabricgoogle
  2. Oct 14, 2026Argon’s wider launch tests whether safety gates can scalegoogle
  3. Oct 14, 2026Argon’s global rollout outruns the evidence on what it can safely dogoogle

Evidence / sources

  1. 1huggingface.co/google/tapnet
All Raw Signals

Next reads

Read next

Open the graph
  1. 01topicGoogleExplore the google topic hub.
  2. 02archiveRaw SignalsOpen every source-linked raw signal.
  3. 03source ledgerSource LedgerInspect the public source evidence graph.
Who read this page?Machine requests, hidden until opened

Loading the privacy-safe route aggregate...

Open the JSON contract