{"slug": "team-of-5-post-trains-glm-5-3-flash-to-fable-5-class-for-300k-all-in", "title": "Team of 5 post-trains GLM 5.3 Flash to Fable 5 class for $300k all-in", "summary": "A five-person team post-trained GLM-5.3 Flash into Ultramarine 1 Flash, an open model it says reaches Fable-class performance on professional knowledge work for $300,000 all-in. Ultramarine 1 Flash scores within three points of Claude Opus 5.5 (max) and GPT-6 Sol (max) on the team's internal PI Knowledge Work benchmark at $0.13 per task, versus $6.96 and $1.26 per task for those models, and generates 6.5% fewer tokens than its GLM-5.3 Flash base. The model is priced at $0.15 per million input tokens, $0.03 cached input, and $0.50 output on the Ultramarine API, the same as GLM-5.3 Flash.", "body_md": "# Ultramarine \n\n1 Flash\n\nAn open model for professional work\n\nUltramarine 1 Flash is Fable-class on knowledge work spanning finance, law, consulting, and operations. It is also 60x cheaper than Fable.\n\n[Technical report →](https://www.paperinstruments.com/blog/ultramarine)\n## Built for professional work\n\n## Improved token efficiency\n\nAcross 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.\n\n## Fable-class model for cents per task\n\nUltramarine 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).\n\n- Input\n- $0.15\n- Cached input\n- $0.03\n- Output\n- $0.50\n\nPer million tokens on the Ultramarine API, the same as GLM-5.3 Flash.\n\nPI 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.\n\nA 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](https://www.paperinstruments.com/blog/ultramarine) for more on the kind of professional data and evaluations that trained this model.\n\n## Crafted with a sense of client readiness\n\nUltramarine 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.", "url": "https://wpnews.pro/news/team-of-5-post-trains-glm-5-3-flash-to-fable-5-class-for-300k-all-in", "canonical_source": "https://www.paperinstruments.com/ultramarine", "published_at": "2026-10-15 17:06:01+00:00", "updated_at": "2026-10-15 17:27:11.146240+00:00", "lang": "en", "topics": ["artificial-intelligence", "large-language-models", "ai-research", "ai-products", "ai-startups"], "entities": ["Ultramarine 1 Flash", "GLM-5.3 Flash", "Claude Fable 5", "GPT-5.6 Sol", "Claude Opus 5.5", "GPT-6 Sol", "PI Knowledge Work", "Artificial Analysis Intelligence Index"], "also_reported_by": [], "alternates": {"html": "https://wpnews.pro/news/team-of-5-post-trains-glm-5-3-flash-to-fable-5-class-for-300k-all-in", "markdown": "https://wpnews.pro/news/team-of-5-post-trains-glm-5-3-flash-to-fable-5-class-for-300k-all-in.md", "text": "https://wpnews.pro/news/team-of-5-post-trains-glm-5-3-flash-to-fable-5-class-for-300k-all-in.txt", "jsonld": "https://wpnews.pro/news/team-of-5-post-trains-glm-5-3-flash-to-fable-5-class-for-300k-all-in.jsonld"}}