d1-3B

We track d1-3B, but the public leaderboard excludes this profile until enough non-generated benchmark coverage is available. Only published rows appear above.

d1-3B is a open weight model with a 32768 tokens context window. No explicit reasoning mode is documented in this profile.

It has 3.12B parameters and accepts text and images within 32,768 tokens. Liquid reports 82.9% mean accuracy on the seven shared text decision tasks and 74.1% across 11 image benchmarks read as decisions, each capped at 1,000 rows. Its reported 8 ms single-question latency uses warm bf16 calls on an RTX 4090 with CUDA graphs, median of 20 runs; it is not token throughput. The weights license has commercial-use conditions.

Liquid AI released d1-3B on October 7, 2026, under the LFM Open License v1.0. It returns Noul, Choice, or Score answers in one forward pass with zero generated tokens. It has 3.12B parameters and accepts text and images within 32,768 tokens. Liquid reports 82.9% mean accuracy on the seven shared text decision tasks and 74.1% across 11 image benchmarks read as decisions, each capped at 1,000 rows. Its reported 8 ms single-question latency uses warm bf16 calls on an RTX 4090 with CUDA graphs, median of 20 runs; it is not token throughput. Liquid AI evaluated these checkpoints with the official Decision Index 0.2.1 scorer on the public split. These are provider-reported runs, not public leaderboard submissions. The index and five areas are skill points, not percent accuracy. They remain separate from Decision Index 0.3, JevBench, and general model rankings; no private vision-split result is reported.

d1-3B sits in the Liquid d1 open family with d1-omni-600M. 6 of 665 tracked benchmark slots currently have displayable evidence. Missing categories stay blank.