ornith-ai·released 2026-08-181 source
Not yet benchmarked
Claims drawn from cited facts, not live model generation.
This model comes from the developer ornith-ai, was released in 2026, and sits at a frontier scale of 403.4 billion parameters. That frontier-scale size implies far more capability headroom and broader task coverage than compact or mid-size models, but it comes at the cost of substantially higher training and inference expense — the classic trade-off where smaller designs win on efficiency and lower cost while larger ones buy capacity and performance.
4 cited facts
This model is not yet benchmarked: it is scored on none of the tracked benchmarks, and it currently holds no top score.
2 cited facts
The average gap to state of the art is not disclosed, so no points-based reading against the leaders can be made under the harness. It holds the top score on 0 benchmarks, which means no category leadership 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
Ornith-1.5-397B is an AI model developed by ornith-ai, released 2026-08-18. Its benchmark record below tags every score as independently reproduced, self-reported, or unverified.