Paper Instruments Releases Ultramarine 1 Flash, an Open Model Claiming Fable-Class Performance at 60x Lower Cost

Performance and Cost Efficiency

Ultramarine 1 Flash, an open model from Paper Instruments, is positioned as a cost-effective alternative for professional knowledge work in finance, law, consulting, and operations. The company claims the model achieves Fable-class performance while being 60x cheaper than Claude Fable 5. On the internal PI Knowledge Work benchmark, Ultramarine 1 Flash scored 59.3, surpassing Claude Fable 5 (57.8) and GPT-5.6 Sol (55.1). The model is priced at $0.15 per million input tokens and $0.50 per million output tokens, identical to GLM-5.3 Flash pricing.

Benchmark Results

The model demonstrates competitive scores across several industry benchmarks, often ranking first or near the top:

• JobBench: 62.9 (1st place, ahead of Gemini 3.8 Flash at 60.6)
• PI Knowledge Work: 59.3 (1st place, ahead of Claude Fable 5 at 57.8)
• FrontierFinance: 53.3 (1st place, ahead of GLM-5.3 Flash at 50.7)
• AutomationBench: 50.2 (2nd place, behind DeepSeek V4.1 Flash at 51.3)
• DiligenceBench: 50.6 (3rd place, behind Gemini 3.8 Flash and Claude Fable 5)
• APEX-Agents 1.1: 57.0 (3rd place, behind Gemini 3.8 Flash and Claude Fable 5)

Token Efficiency and Availability

Ultramarine 1 Flash generates 6.5% fewer tokens than its base model, GLM-5.3 Flash, across the six benchmarks tested. The model is available via the Ultramarine API, with model weights announced as "coming soon" on GitHub. The company highlights the model's ability to produce clean, professional deliverables such as Excel workbooks, PowerPoint decks, and Word documents, adhering to strict visual and information hierarchy standards.