CMSManhattan·released 2026-08-021 source
Not yet benchmarked
Claims drawn from cited facts, not live model generation.
Developed by CMSManhattan, this model is a compact 1.78-billion-parameter release from 2026, a scale that favors efficiency and lower cost rather than the greater headroom and expense of larger systems.
3 cited facts
This model is not yet benchmarked, with zero tracked benchmarks on which it is scored. Accordingly, it holds no current top score among those benchmarks.
2 cited facts
Average SOTA gap is unknown for this model, and it holds the top score on 0 benchmarks, meaning there is no evidence of category leadership on any of its reported benchmarks yet.
2 cited facts
2 facts cross-checked across data sources: 2 single-source.
Source facts, citations, and refresh stamp for this record.
Sources: huggingface_models
JiRackUltra_1b is an AI model developed by CMSManhattan, released 2026-08-02. Its benchmark record below tags every score as independently reproduced, self-reported, or unverified.