From the source
Subquadratic released SubQ Snap v1.1, an update to its frontier decision model that it says is the smartest and most capable decision model in production, built for long-context reasoning and agentic decisions.
- The company says Snap v1.1 ties for first on the Artificial Analysis Decision Index with a score of 65, and that it leads Jev 1.14 by 7.2 points and Snap 1.0 by 5.5 across 16 enterprise benchmarks.
- On When2Call, which tests whether an agent should call a tool, ask for missing information or decline, the model rises from 75.2 to 96.2.
- Subquadratic says post-training that calibrates uncertainty cut expected calibration error on held-out problem families from 0.108 to about 0.014, while overconfident hallucinations on out-of-distribution queries fell from 99.3% to 12.5%.
- Snap 1.1 ships with 384K tokens of context, and the company cut its price to $0.04 per 1M tokens, a 20% reduction.