Apertus 1.5 release notes: 19T/17T total training tokens, fully open weights, data and recipes
lmoroney · x · 2026-10-17
Per the official Hugging Face release notes, Apertus 1.5 is continued pretraining of the first-generation models: 4 trillion additional tokens for 8B and 2 trillion for 70B, totaling 19T and 17T respectively.
Key enhancements: native multimodal understanding (documents, diagrams, photos; experimental speech), optional thinking mode, 262,144-token context, improved instruction-following and tool use, plus stronger adherence to the Apertus Charter. Fully open: weights, data, and complete training details.
Related event: Apertus 1.5 Open-Source Release Adds Image, Audio and 262K Context(2 posts)→
More from Models
- Open-Weights Frontier Gap Narrows to Under 3 Months as DS-Flash Beats GPT-6.1 Sol — teortaxesTex · 2026-10-17
- DeepSeek-V4.1-Flash-1017 Update Tops Open-Weights, Beats Closed Models on Several Agent Benchmarks — teortaxesTex · 2026-10-17
- DeepSeek post-training lead Wu Yu declares the era of democratized wish-driven coding — teortaxesTex · 2026-10-17
- DeepSeek Quietly Uploads Multimodal DeepSeek-V4.1-Flash-1017 to Hugging Face — scaling01 · 2026-10-17
- Pooling OpenAI subscriptions to bypass regional inference capacity limits — geoffreyhuntley · 2026-10-17
- DeepSeek-V4.1-Flash-1017 repo appears on Hugging Face, hinting at a new release — Bestlife73 · 2026-10-17