Paper Instruments launches Ultramarine 1 Flash for professional work
CEO Daanish Khazi says post-training cost about $300,000; the company's benchmarks put one task at $0.13, with model weights expected in about two weeks.
By Ryan Merket · Published
Primary source: X
Why it matters
Paper Instruments is testing whether a small team can use reinforcement learning on an open base model to make professional-work agents cheaper to run. Its strongest cost comparison comes from an internal benchmark, so the open weights and outside evaluations will be important tests of the claim.
Paper Instruments CEO Daanish Khazi (@bertgodel) launched Ultramarine 1 Flash on October 14th, pitching it as an open model for professional work across finance, law, consulting and operations. In a thread on X, Khazi said a five-person team post-trained the model for about $300,000 and claimed it performs similarly to Anthropic's Claude Fable 5 at roughly one-sixtieth the cost.…