ModelsAgents

Paper Instruments releases Ultramarine 1 Flash for knowledge-work agents

Paper Instruments ·

Paper Instruments technical report cover reading "Ultramarine 1 Flash: Scaling Non-Verifiable RL" beside a sweep of ultramarine-blue paint

Ultramarine 1 Flash is an agentic model for finance, law, consulting and operations work, built by applying reinforcement learning to GLM-5.3 Flash. Paper Instruments says it outscores Claude Fable 5 on its own internal suite ($0.13 vs $8.19 per task). The company documents a paid Ultramarine API and says model weights are "coming soon"; that is the company's promise, with no date given.

How you’d use it: Compare a cheaper knowledge-work model on your own memo or diligence tasks before relying on vendor benchmarks.

Sources

← All AI news

Get the daily AI news

The day's important AI news by email: short, plain summaries, each linked to its source. Unsubscribe any time.

Prefer email? Write to [email protected] ([email protected]).