Ultramarine 1 Flash 发布，以 Fable 60% 成本登顶知识工作基准

Ultramarine 1 Flash is Fable-class on knowledge work spanning finance, law, consulting, and operations. It is also 60x cheaper than Fable.

Ultramarine API →Model weights coming soon

Built for professional work

JobBench

1. Ultramarine 1 Flash62.9
2. Gemini 3.8 Flash60.6
3. DeepSeek V4.1 Flash60.4
4. Claude Fable 557.4
5. GLM-5.3 Flash56.3
6. Kimi K354.3
7. GPT-5.6 Sol45.4

PI Knowledge Work (internal)

1. Ultramarine 1 Flash59.3
2. Claude Fable 557.8
3. GPT-5.6 Sol55.1
4. GLM-5.3 Flash52.0
5. Kimi K351.2
6. Gemini 3.8 Flash49.4
7. DeepSeek V4.1 Flash47.9

FrontierFinance

1. Ultramarine 1 Flash53.3
2. GLM-5.3 Flash50.7
3. Gemini 3.8 Flash50.0
4. Claude Fable 549.2
5. GPT-5.6 Sol46.8
6. Kimi K346.4
7. DeepSeek V4.1 Flash46.0

AutomationBench

1. DeepSeek V4.1 Flash51.3
2. Ultramarine 1 Flash50.2
3. GLM-5.3 Flash48.8
4. Gemini 3.8 Flash47.3
5. Kimi K346.7
6. Claude Fable 546.2
7. GPT-5.6 Sol45.8

DiligenceBench

1. Gemini 3.8 Flash53.1
2. Claude Fable 552.9
3. Ultramarine 1 Flash50.6
4. Kimi K350.4
5. DeepSeek V4.1 Flash50.3
6. GPT-5.6 Sol50.0
7. GLM-5.3 Flash48.2

APEX-Agents 1.1

1. Gemini 3.8 Flash64.3
2. Claude Fable 563.6
3. Ultramarine 1 Flash57.0
4. GLM-5.3 Flash52.8
5. GPT-5.6 Sol51.5
6. Kimi K350.6
7. DeepSeek V4.1 Flash39.5

Improved token efficiency

Across the six benchmarks, Ultramarine 1 Flash generates 6.5% fewer tokens than its base, GLM-5.3 Flash, taking a shorter path to better work.

GLM-5.3 Flash32.7MUltramarine 1 Flash30.5M−6.5%Output tokens generated across all six benchmarks, reasoning included. Where GLM-5.3 Flash’s score is a public one, its tokens come from our own runs closest to that score.

Fable-class model for cents per task

Ultramarine 1 Flash sits on the cost-quality frontier at the most attractive price point. At $0.13 per task, it outscores Claude Fable 5 (max) and GPT-5.6 Sol (max), at$8.19 and $2.09 per task, and comes within three points of Claude Opus 5.5 (max) and GPT-6 Sol (max) at$6.96 and $1.26 per task. We measure this on PI Knowledge Work, a carefully built internal benchmark whose scores correlate strongly with the Artificial Analysis Intelligence Index (r = 0.81).

Input$0.15Cached input$0.03Output$0.50

Per million tokens on the Ultramarine API, the same as GLM-5.3 Flash.

PI Knowledge Work is our internal benchmark of professional work in finance, law, consulting, and operations. Its tasks ask the model to deliver finished Excel workbooks, PowerPoint decks, and Word documents, and to work inside live business systems such as ERP and MES.

A rubric written for each task scores the result on substance, accuracy, and presentation. Every model works in the same harness, and costs are at each developer's official API pricing. Read our technical report for more on the kind of professional data and evaluations that trained this model.

Crafted with a sense of client readiness

Ultramarine 1 Flash writes clean, simple, straightforward prose and has a strong sense of visual and information hierarchy. It is built to adhere to the same rules required of analysts in professional services.

Nucor through the steel cycle

Five periods of segment sales and earnings in one table, with every subtotal and total reconciling. Steel mills earnings fell from $7.2bn in 2022 to$2.4bn in 2025, while their share of sales rose from 58% to 62%.

---

作者发帖说明（HN）：

https://www.paperinstruments.com/blog/ultramarine

https://twitter.com/bertgodel/status/2110400987622265232