TL;DR
Get audio and creator gear delivered free with Prime
- Fast, free delivery on millions of items
- Prime Video, Amazon Music and more included
- Member-only deals all year
Paper Instruments has introduced Ultramarine 1 Flash, an open model aimed at finance, legal, consulting and operations work. The company reports a score of 59.3 on its internal PI Knowledge Work benchmark for an estimated $0.13 per task, but the result is based on the company’s own evaluation and has not been independently verified in the supplied material.
Paper Instruments has introduced Ultramarine 1 Flash, an open model designed for professional tasks in finance, law, consulting and operations, and says it scored 59.3 on its PI Knowledge Work benchmark at an estimated $0.13 per task. The release matters to businesses weighing the cost of AI-assisted work, but the performance figures come from the company’s own benchmark rather than an independent evaluation described in the supplied report.
Paper Instruments describes Ultramarine 1 Flash as a model for producing work such as Excel workbooks, PowerPoint presentations and Word documents, as well as operating within business systems including ERP and MES software. Its benchmark tasks cover finance, law, consulting and operations. The company says each task is scored against a task-specific rubric for substance, accuracy and presentation, with models run in the same evaluation harness.
In the comparison published by Paper Instruments, Ultramarine 1 Flash with maximum reasoning scored 59.3 for $0.13 per task. The report lists Claude Fable 5 at maximum reasoning at 57.8 for $8.19 per task, and GPT-5.6 Sol at maximum reasoning at 55.1 for $2.09. It also places the model within three points of Claude Opus 5.5 at maximum reasoning, scored 61.9 at $6.96, and GPT-6 Sol at maximum reasoning, scored 61.4 at $1.26.
The listed Ultramarine API rates are $0.15 per million input tokens, $0.03 per million cached input tokens and $0.50 per million output tokens. Paper Instruments says these rates match those of GLM-5.3 Flash. It also reports that Ultramarine generated 6.5% fewer tokens than GLM-5.3 Flash across six benchmarks; the company says its token counts for that comparison came from its own runs chosen to come closest to GLM’s public score.
A Lower-Cost Route to Business Tasks
If the company’s benchmark results transfer to real work, the model’s reported task cost and score could make AI-assisted document and spreadsheet production more affordable for organizations that handle repeated professional workflows. The comparison is framed around completed deliverables—not just text generation—so quality in accuracy, substance and presentation is central to whether the lower price is useful.
However, the figures should be read as Paper Instruments’ reported results, not as a guarantee of savings or performance in a buyer’s own systems. The supplied material does not quantify deployment expenses, human review, error rates in production or the range of tasks for which the model performs reliably. Those factors can affect the actual cost and value of using it.
As an affiliate, we earn on qualifying purchases.
How Paper Instruments Tested the Model
The company’s PI Knowledge Work benchmark is described as an internal set of tasks spanning finance, law, consulting and operations. It asks models to create finished files and complete work inside live business software. Paper Instruments says its scoring rubrics assess substance, accuracy and presentation, and that models use the same harness while costs are calculated from each developer’s official API pricing.
Paper Instruments says PI Knowledge Work scores correlate with the Artificial Analysis Intelligence Index, reporting a correlation of r = 0.81. That is a claim about the relationship between its internal benchmark and another index; it does not make the benchmark an independent test. The source also says its technical report discusses the professional data and evaluations used to train the model, but the supplied excerpts do not provide detailed methods or independent replication.
“Ultramarine 1 Flash is Fable-class on knowledge work spanning finance, law, consulting, and operations.”
— Paper Instruments
As an affiliate, we earn on qualifying purchases.
Independent Results Remain Unreported
The supplied report does not identify an independent evaluation of Ultramarine 1 Flash or show results from customers using it in production. It also does not include enough detail here to assess how representative the benchmark tasks are of different organizations’ work, how much human intervention was needed, or how scores vary across individual tasks.
The headline cost comparison depends on the benchmark’s definition of a task and the pricing assumptions used. The source does not clarify whether the $0.13 estimate includes all operational costs beyond model API usage, such as review, integration, or repeat attempts. The launch date, access terms, licensing details and specific model-weight availability are also not established in the supplied material, despite the page describing the model as open.
As an affiliate, we earn on qualifying purchases.
Technical Details and Access Terms
Paper Instruments points readers to a technical report, Ultramarine API and GitHub for further information. The supplied source does not specify a release schedule, model license, hardware requirements or whether the complete model weights are available for download. Those details will help developers and businesses determine what “open” means for this release and whether the model can be tested outside the company’s API.
The next useful evidence will be independent benchmark results and documented use in real business workflows. Until then, readers can treat the published score and task costs as company-reported comparisons, and should test the model against their own requirements before relying on it for professional deliverables.
AI-powered Excel and PowerPoint tools
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Key Questions
What is Ultramarine 1 Flash?
Paper Instruments describes it as an open model for professional work, including tasks in finance, law, consulting and operations. The company says it can produce finished office documents and work within ERP and MES systems.
What score did Paper Instruments report?
The company reports a score of 59.3 at maximum reasoning on its internal PI Knowledge Work benchmark, with an estimated cost of $0.13 per task.
Are the benchmark results independently verified?
The supplied report does not describe an independent verification. The PI Knowledge Work benchmark is identified as Paper Instruments’ internal evaluation, so its comparisons should be treated as company-reported results.
How much does API use cost?
Paper Instruments lists rates of $0.15 per million input tokens, $0.03 per million cached input tokens and $0.50 per million output tokens. The separate $0.13-per-task figure is an estimate based on its benchmark.
Where can developers find more information?
Paper Instruments links to a technical report, its Ultramarine API and GitHub. The supplied material does not spell out the model’s license or exact availability terms.
Source: hn
Halloween Picks
halloween
As an affiliate, we earn on qualifying purchases.
