ClassyFire PubChem Labels (2026-04-11)

Open dataset of recovered ClassyFire labels for PubChem compounds. This snapshot contains 257112 successful ClassyFire responses exported as a single merged JSONL.zst file. Rows that ended as misses (26), invalid input (0), or failed requests (1470) are excluded from the release artifact. The release contains a merged classyfire-labels.jsonl.zst dataset and a machine-readable manifest.json describing the published snapshot. Handled rows so far: 258608. Successful rows: 257112. Misses: 26. Invalid rows: 0. Failed rows: 1470. Format: JSON Lines compressed with Zstandard. Updated periodically from the streaming crawler. Source code: classyfire.