ynklab/Tower-7B-Sep_0c1t

TEXT GENERATIONPricing:Input $0.4 / Cached $0.08 / Output $0.8Concurrent Unit Cost:1Model Size:7BQuant:FP8Context Size:4kTool Calling:SupportedPublished:Sep 7, 2026License:cc-by-nc-sa-4.0Architecture:Transformer Open Weights Featherless Exclusive Cold

ynklab/Tower-7B-Sep_0c1t is a 7 billion parameter language model, fine-tuned by ynklab from Unbabel/TowerInstruct-Mistral-7B-v0.2, specifically for multilingual chunk-level machine translation. This variant, 'Sep_0c1t', translates individual source chunks into target chunks without relying on preceding context. It supports translation between English and nine other languages, optimized for document-level translation tasks using fixed-range chunking.

Loading preview...

Model Overview

Tower-7B-Sep_0c1t is a 7 billion parameter language model developed by ynklab, based on the Unbabel/TowerInstruct-Mistral-7B-v0.2 architecture. It is specifically fine-tuned for multilingual chunk-level machine translation, a method designed for length-consistent document translation. This model is the Sep_0c1t variant, meaning it translates single source chunks into single target chunks independently, without incorporating prior contextual chunks.

Key Capabilities

  • Multilingual Translation: Supports translation between English and German, Spanish, French, Italian, Korean, Dutch, Portuguese, Russian, and Chinese in both directions.
  • Chunk-Level Translation: Optimized for translating text in fixed-range chunks (256–512 tokens), as detailed in the associated Doc2FRC paper.
  • Document-Level Consistency: Designed to be integrated into a larger system for length-consistent document translation, with specific inference scripts available in the Doc2FRC GitHub repository.

Training Details

The model underwent full-parameter supervised fine-tuning for 2 epochs, utilizing the sardinelab/DocBlocks dataset. It was trained with a maximum sequence length of 32,768 tokens and a ChatML-style prompt format for translation tasks.

Recommended Use Cases

This model is ideal for researchers and developers working on document-level machine translation, particularly those interested in chunk-based translation approaches. For optimal performance in document translation, users should refer to the specific chunking and inference procedures outlined in the Doc2FRC repository.

Finetunes
1 models