Live drop
Ultramarine 1 Flash
Paper InstrumentsDropped Oct 14, 2026
Post-trained GLM-5.3 Flash for professional knowledge work that matches frontier models at ~60x lower per-task cost.
About this release
Live drop
Paper InstrumentsDropped Oct 14, 2026
Post-trained GLM-5.3 Flash for professional knowledge work that matches frontier models at ~60x lower per-task cost.
About this release
Ultramarine 1 Flash is a post-trained version of GLM-5.3 Flash using large-scale non-verifiable RL on synthetic professional tasks across finance, law, consulting, and operations. It improves on the base across six knowledge-work benchmarks with fewer output tokens and is available now via an OpenAI-compatible API at the same rates as the base model; weights are scheduled for release soon. The model targets document-heavy workflows like memos, models, and decks.
Read the official Paper Instruments launch pageFirst builds incoming — this lane fills in automatically as people start building with Ultramarine 1 Flash. Know one? Submit the URL.