GLM 5.2 Released
Hacker News (AI keywords)14 hours agoReleaseZhipu AI has released GLM 5.2, a point update to its flagship General Language Model series. GLM models are widely used for multilingual tasks, particularly in Chinese-language applications, and are available both as a commercial API and as open-weight downloads. The release was noted on Hacker News, though specific feature changes, benchmark results, and technical details for version 5.2 were not available from the source.
Can Voice Agents Handle Bilingual Customers? Benchmarking Frontier ASR on Code-Switched Speech
Hugging Face Blog4 days agoBenchmarkCode-switching—where bilingual speakers blend two languages in a single utterance—is common in markets like Taiwan, Singapore, and India, yet most ASR benchmarks focus on monolingual audio. ServiceNow AI evaluates frontier speech recognition models specifically on this mixed-language scenario. The findings help enterprise teams make informed ASR model choices when deploying voice agents for multilingual customer-facing applications.
Cohere and Mila Partner to Advance Quebec French Language and Culture in AI
Cohere Blog6 days agoBusinessCohere has partnered with Mila, the Quebec AI Institute, to improve the representation of Quebec French (Québécois) and its cultural nuances in AI. The collaboration aims to address the European French bias in current models by leveraging Cohere's multilingual capabilities and Mila's research expertise. This initiative will help deliver more culturally accurate AI solutions for Quebec's public and private sectors.
Introducing Cohere Transcribe: A New State-of-the-Art in Open-Source Speech Recognition★ 80
Cohere Blog6 days agoReleaseCohere has announced "Cohere Transcribe," a new state-of-the-art open-source speech recognition model. Designed to deliver highly accurate and efficient speech-to-text capabilities, it represents Cohere's expansion into open-source audio AI. The model aims to challenge existing industry benchmarks like OpenAI's Whisper by offering superior multilingual performance.
Cohere Blog: Technology Tag Page Overview
Cohere Blog6 days agoCommentaryThis page aggregates all technology-focused articles on the Cohere blog. As an enterprise-focused AI company, Cohere's technical content primarily covers its Command LLM family, industry-leading Embed and Rerank models, and practical RAG implementation guides. It serves as a key resource for developers and enterprise architects tracking Cohere's technical evolution.
Cohere's Commitment to Open Science and Collaborative AI Research
Cohere Blog6 days agoCommentaryCohere's Open Science initiative, primarily driven by its non-profit research lab Cohere For AI (C4AI), focuses on democratizing AI research. By releasing open-weights models like Aya and fostering global research collaborations, Cohere aims to bridge the gap in multilingual AI representation. This approach highlights their commitment to community-driven, accessible AI development.
Cohere Official Research Blog and Technical Publications
Cohere Blog6 days agoCommentaryThe Cohere Research blog serves as the central hub for the company's academic papers and technical breakthroughs. It covers key areas including advanced Retrieval-Augmented Generation (RAG), multilingual embeddings, and robust tool-use capabilities for enterprise agents. This is a key resource for understanding the foundational technology behind Cohere's models.
Voxtral★ 78
Mistral AI News6 days agoReleaseMistral AI introduces Voxtral, a speech understanding model family with 24B and 3B variants under Apache 2.0. The models support long-context transcription, audio Q&A, summarization, multilingual detection, and function calling from voice. Mistral says Voxtral is competitive across transcription and audio understanding benchmarks, with API access starting at $0.001 per minute and local downloads available on Hugging Face.
Voxtral TTS: Open-Weights, Low-Latency Text-to-Speech from Mistral AI★ 78
Mistral AI News6 days agoReleaseMistral AI introduced Voxtral TTS, its first text-to-speech model, focused on realistic multilingual voice generation. The 4B-parameter model supports nine languages, quick voice adaptation from short references, and low-latency streaming for voice agents. Mistral says human evaluations show stronger naturalness than ElevenLabs Flash v2.5, with API access, Studio testing, Le Chat access, and open weights on Hugging Face.
Voxtral TTS★ 76
Mistral AI News6 days agoReleaseMistral AI introduced Voxtral TTS, its first text-to-speech model, targeting natural multilingual voice generation across nine languages. The 4B-parameter model supports voice adaptation from short references, emotional expressiveness, dialect handling, and low-latency streaming. It is available through API, Mistral Studio, and Le Chat, with open weights on Hugging Face under a non-commercial CC BY NC 4.0 license.
Scaling multilingual diplomacy during the Polish presidency of the Council of the EU
ElevenLabs Blog6 days agoBusinessBased only on the title, this ElevenLabs Blog post likely discusses multilingual diplomacy during Poland’s presidency of the Council of the EU. It may involve voice, translation, or audio workflows, but the original text is unavailable, so specific claims cannot be verified. The main signal is that AI voice tools are being positioned for public-sector and international communication use cases.
IBM 發布 Granite Embedding Multilingual R2：具備 32K 上下文與 Apache 2.0 授權，100M 參數以下最強多語言嵌入模型★ 75
Hugging Face Blog30 days agoReleaseIBM has officially released a new multilingual embedding model on the Hugging Face platform called "Granite Embedding Multilingual R2." The model's most…
開源 AI 資源週報 (#20)：全新組織與模型類型登場！涵蓋 Nemotron Super、Sarvam、Cohere Transcribe 等最新進展
Interconnects (Nathan L.)76 days agoReleaseProminent AI scholar and commentator Nathan Lambert, in his latest edition of Latest Open Artifacts (#20), has compiled the major recent developments in the…
Hugging Face 推出 Open ASR Leaderboard 新賽道：聚焦多語言與長音訊語音辨識趨勢★ 75
Hugging Face Blog205 days agoReleaseHugging Face recently made a major upgrade to its flagship "Open ASR Leaderboard," officially launching two brand-new evaluation tracks: "Multilingual" and…
mmBERT：ModernBERT 邁向多語言時代，開源高效能多語言編碼器模型登場★ 78
Hugging Face Blog278 days agoReleaseIn today's era dominated by generative AI and large language models (LLMs), bidirectional encoder models (such as BERT and RoBERTa) still play an indispensable…
NVIDIA 於 Hugging Face 開源發布 600 萬筆多語言推理數據集★ 78
Hugging Face Blog297 days agoReleaseNVIDIA has officially released a massive "Multi-Lingual Reasoning Dataset" containing 6 million samples on the Hugging Face platform. This significant…
Hugging Face 發表 SmolLM3：輕量、多語言、長上下文的端側推理模型★ 80
Hugging Face Blog341 days agoReleaseHugging Face has announced the release of a brand-new generation of lightweight open-source models — SmolLM3. As the latest member of the SmolLM family…
Visual Salamandra 7B 發布：巴塞隆納超級電腦中心推出開源多模態大模型，主打多語言與視覺理解★ 70
Hugging Face Blog429 days agoReleaseThe Language Technologies department (BSC-LT) of the Barcelona Supercomputing Center (BSC) recently released a new open-source multimodal model on Hugging Face…
Hugging Face 推出阿拉伯語 LLM 評估新標準：引入阿拉伯語指令遵循（IFEval）與更新 AraGen
Hugging Face Blog432 days agoReleaseHugging Face recently announced a major upgrade to its Arabic Large Language Model (LLM) leaderboard, aiming to provide a more credible and comprehensive…
Google 推出全新 Gemma 3：支援多模態、多語言與長文本的開源大語言模型★ 90
Hugging Face Blog459 days agoReleaseGoogle has officially launched Gemma 3, the next generation of its open-source large language model series — a major technical leap forward from Gemma 2. Gemma…
深入解析 Aya Vision：推動多語言多模態 AI 的前沿發展★ 75
Hugging Face Blog467 days agoReleaseCohere For AI (C4AI) has officially launched "Aya Vision," a series of open-source multimodal models (available in 8B and 32B parameter versions) designed…
Hugging Face 與印度科學理工學院（IISc）達成合作，加速印度多元語言的 AI 模型開發
Hugging Face Blog472 days agoBusinessHugging Face has announced a formal partnership with India's premier academic institution — the Indian Institute of Science (IISc) — with the core goal of…
Google 推出 SigLIP 2：更強大的多語言視覺語言編碼器★ 80
Hugging Face Blog478 days agoReleaseGoogle has officially launched SigLIP 2, a major upgrade to its widely popular SigLIP (Sigmoid Loss for Language-Image Pre-training) vision-language encoder…
視覺文件檢索邁向多語言：Hugging Face 推出 VDR-2B-multilingual 模型★ 80
Hugging Face Blog520 days agoReleaseHugging Face has recently released a new Visual Document Retrieval (VDR) model — **VDR-2B-multilingual**. This technology marks a formal transition in document…
讓大型模型展開辯論：首屆多語言 LLM 辯論賽★ 75
Hugging Face Blog571 days agoReleaseThis article from the Hugging Face blog introduces "The First Multilingual LLM Debate Competition." As large language models (LLMs) have rapidly advanced…
深入解析 Aya Expanse：推動多語言 AI 模型的前沿進展★ 75
Hugging Face Blog598 days agoReleaseCohere For AI (C4AI) has officially launched Aya Expanse, a family of open-weight models designed specifically for multilingual tasks. The family includes two…
🇨🇿 BenCzechMark：你的 LLM 能聽懂捷克語嗎？全新捷克語基準測試發布
Hugging Face Blog621 days agoReleaseThe Hugging Face team and its collaborators have jointly launched a new benchmark called "BenCzechMark," designed to evaluate the understanding and generation…
Hugging Face 推出 Open Arabic LLM 排行榜，加速阿拉伯語大語言模型評測與發展
Hugging Face Blog761 days agoReleaseHugging Face has announced the launch of the "Open Arabic LLM Leaderboard," an important initiative aimed at advancing Arabic natural language processing (NLP)…
Hugging Face 推出希伯來語 LLM 開放排行榜，推動非英語系 AI 模型評測
Hugging Face Blog770 days agoReleaseHugging Face has officially launched the "Open Leaderboard for Hebrew LLMs," an open-source evaluation platform specifically designed for Hebrew large language…
Huggy Lingo：利用機器學習改善 Hugging Face Hub 上的語言元數據 (Metadata)
Hugging Face Blog1,047 days agoReleaseHugging Face Hub, the world's largest open-source AI community platform, hosts hundreds of thousands of models, datasets, and demo applications (Spaces). For a…