Model weights from a from-scratch replication of TabPFN v1 (Hollmann et al., ICLR 2023) in JAX + optax (no flax). Prior-data-fitted networks that do in-context learning over rows of a tabular dataset — no gradient updates at predict time.

These are params-only checkpoints (optimizer state stripped). Each is a Python pickle of {"params": <nested dict>, "cfg": <ModelConfig kwargs>, "step": int}.

Evaluated under the TabPFN v1 paper's protocol — 18 numerical OpenML-CC18 datasets (≤1,000 training points), 5× 50/50 splits, ROC AUC one-vs-one macro.

v1_d matches the published v1 result to within the paper's own ±0.0086 std.

Parity is on small, clean, numerical classification (v1's home turf, where v1 and v2 are within 0.002 of each other). On the larger, messier TabArena-Lite benchmark these models trail TabPFN v2 (0.875) by ~0.03 — a gap that scales with dataset size and traces to training context length (T=1024), not the prior.

See manifest.json for each model's prior, metrics, and parameter count.