深入解析 Aya Vision：推動多語言多模態 AI 的前沿發展
Original: A Deepdive into Aya Vision: Advancing the Frontier of Multilingual Multimodality
Cohere For AI (C4AI) has officially launched "Aya Vision," a series of open-source multimodal models (available in 8B and 32B parameter…
Cohere For AI 發表全新開源多語言多模態模型「Aya Vision」（提供 8B 與 32B 版本）。該模型旨在解決現有視覺語言模型（VLM）嚴重偏向英語的問題，大幅提升了對全球多種語言（特別是低資源語言）的圖像理解與文字生成能力。在多項多語言多模態基準測試中，Aya Vision 展現出超越同量級開源模型的優異性能，並已於 Hugging Face 平台上開源。
Cohere For AI (C4AI) has officially launched "Aya Vision," a series of open-source multimodal models (available in 8B and 32B parameter versions) designed specifically for multilingual environments, with the goal of breaking through the limitations of current vision-language models (VLMs), which are predominantly English-centric.
Free shows the 3-line summary; Pro unlocks the full deep summary (~300 words) so you never have to click through.
See Pro plans →Want the original English / full article?
Read on Hugging Face Blog →Summaries are AI-generated; the original article is authoritative.