{
  "$type": "site.standard.document",
  "bskyPostRef": {
    "cid": "bafyreic5j7z5jeuedysp2cjepammwrjlh5ugaxihhh2vq2zyjvwlvaeac4",
    "uri": "at://did:plc:l5eubmmg57bpj76f4frmgy3r/app.bsky.feed.post/3mp6oo6quj7v2"
  },
  "coverImage": {
    "$type": "blob",
    "ref": {
      "$link": "bafkreicuixqm3c72ognwxc4gwo4mdesov2po5exhd2nc2km7oc3svuxweu"
    },
    "mimeType": "image/png",
    "size": 55821
  },
  "path": "/blog/hybrid-token-prediction",
  "publishedAt": "2026-06-25T08:00:00.000Z",
  "site": "https://allenai.org",
  "textContent": "New token-level analyses of Olmo 3 and Olmo Hybrid show that hybrid models predict meaning-bearing, context-dependent tokens better than transformers, while transformers retain an edge on verbatim copying.",
  "title": "Which tokens does a hybrid model predict better?"
}