TL;DR
Get art and craft supplies 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 professional work in finance, law, consulting and operations. The company reports a 59.3 score on its internal PI Knowledge Work benchmark at an estimated $0.13 per task, while details needed to independently assess the comparison remain limited in the supplied material.
Paper Instruments has introduced Ultramarine 1 Flash, an open model designed for professional tasks in finance, law, consulting and operations. The company says it scored 59.3 on its internal PI Knowledge Work benchmark at an estimated $0.13 per task, positioning it as a lower-cost option for producing business documents and working in enterprise systems.
The benchmark covers tasks that ask models to produce finished Excel workbooks, PowerPoint decks and Word documents, as well as work inside live business systems such as ERP and manufacturing execution systems. Paper Instruments says a task-specific rubric scores substance, accuracy and presentation, and that models use the same evaluation harness. The company says its benchmark scores correlate with the Artificial Analysis Intelligence Index at r = 0.81; the supplied material does not include an independent audit of that relationship.
Paper Instruments reports that Ultramarine 1 Flash generated 6.5% fewer tokens than its base model, GLM-5.3 Flash, across six benchmarks. The comparison counts output tokens, including reasoning. The company says GLM-5.3 Flash’s scores are public, while its token counts come from company runs selected to be closest to those scores.
The company lists API rates of $0.15 per million input tokens, $0.03 for cached input and $0.50 for output, matching the rates it gives for GLM-5.3 Flash. Its task-cost comparison estimates $0.13 per PI Knowledge Work task for Ultramarine 1 Flash. Paper Instruments reports scores and costs of 57.8 and $8.19 for Claude Fable 5 at maximum reasoning, and 55.1 and $2.09 for GPT-5.6 Sol at maximum reasoning. These are company-reported benchmark results, not an independent assessment of general performance.
A Lower Cost for Document Work
If the reported results hold across typical business workloads, the model’s estimated per-task cost could make repeated document and operations tasks less expensive to run. That matters to organizations weighing whether to automate work that requires a finished spreadsheet, presentation or report, rather than a short text response.
The result also highlights a practical trade-off for buyers: a benchmark score and token price do not establish how a model performs on a particular company’s data, workflow or quality standards. The reported $0.13 task estimate is tied to Paper Instruments’ own benchmark and pricing assumptions. Organizations would need to compare output quality, integration needs and total usage costs on their own workloads before relying on it.
professional AI document automation software
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
How Paper Instruments Measured It
PI Knowledge Work is described by Paper Instruments as an internal benchmark of professional tasks across four fields: finance, law, consulting and operations. Tasks require completed artifacts or actions in business software. A rubric written for each task scores the result on substance, accuracy and presentation.
Paper Instruments says all models run in the same harness and that task costs use each developer’s official API pricing. Its product page presents comparisons at different reasoning settings, including low, medium, high and maximum for some models. The supplied material does not state the benchmark’s task count, full evaluation dates, uncertainty ranges or whether model developers reviewed the results.
The page also cites an example involving Nucor: a table covering five periods of segment sales and earnings, with the company saying all subtotals and totals reconcile. It reports steel mills earnings falling from $7.2 billion in 2022 to $2.4 billion in 2025, while that segment’s share of sales rose from 58% to 62%. The example illustrates the type of business analysis the model is intended to produce; it does not establish performance across all tasks.
““Ultramarine 1 Flash is Fable-class on knowledge work spanning finance, law, consulting, and operations.””
— Paper Instruments
enterprise AI language model for finance law consulting
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Questions About the Benchmark
The supplied product-page material does not state when the evaluations were run, how many tasks were included, or how much scores varied between runs. It also does not provide task-level results or independent replication, leaving readers unable to assess the full benchmark from the figures alone.
The meaning of “open model” is not detailed in the material: it does not specify the model’s license, whether weights are downloadable, or what restrictions apply. The page lists API prices but does not clarify availability, rate limits, regional access or whether the reported task cost includes all operational expenses. Those details affect whether the reported cost and openness claims apply to a buyer’s intended use.
AI-powered business report generator
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Availability and Independent Results
Paper Instruments points readers to its Ultramarine API and technical report, as well as a GitHub link. The supplied source does not specify a release date, model-weight access terms or a schedule for further evaluations.
For buyers, the next useful evidence would include the full technical report, task-level benchmark results, licensing and deployment details, and tests on representative company workflows. Until those are available, the published score and price comparisons should be treated as Paper Instruments’ reported results.
AI tools for enterprise document creation
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Key Questions
What is Ultramarine 1 Flash?
It is an open model announced by Paper Instruments for professional work in finance, law, consulting and operations. The supplied material does not define its license or explain whether its weights can be downloaded.
What score did it receive?
Paper Instruments reports a score of 59.3 on its internal PI Knowledge Work benchmark at maximum reasoning. The source does not provide the task count or independent validation.
How much does a task cost?
The company estimates $0.13 per task on PI Knowledge Work. Its listed API rates are $0.15 per million input tokens, $0.03 per million cached input tokens and $0.50 per million output tokens.
What work does the benchmark cover?
Paper Instruments says tasks include producing finished spreadsheets, presentations and documents, and working inside systems such as ERP and manufacturing execution software.
Are the performance claims independently confirmed?
The supplied source is Paper Instruments’ own product material. It describes the company’s benchmark and comparisons but does not include independent replication or audit results.
Source: hn
Fall Picks
fall essentials
As an affiliate, we earn on qualifying purchases.
