{"slug":"ovh-ai-olala-mini-7a1b-c13b50ec","url":"https://silinex.xyz/wire/ovh-ai-olala-mini-7a1b-c13b50ec","json_url":"https://silinex.xyz/wire/ovh-ai-olala-mini-7a1b-c13b50ec.json","md_url":"https://silinex.xyz/wire/ovh-ai-olala-mini-7a1b-c13b50ec.md","headline":"European cloud services company OVH enters the AI model space, open-sourcing Olala mini 7A1B.","headline_zh":"欧洲云服务企业 OVH 涉足 AI 模型，开源发布 Olala mini 7A1B","title":"European cloud services company OVH enters the AI model space, open-sourcing Olala mini 7A1B.","summary":"Based on public reports including IT之家: European cloud service provider OVHcloud announced today local time the launch of its first AI model family, Olala, along with the first member of the series, Olala mini 7A1B.","summary_zh":"综合 IT之家 等公开报道整理：欧洲云服务企业 OVHcloud 当地时间今日宣布推出其首个 AI 模型家族 Olala，同时公布了该系列的首个成员 Olala mini 7A1B。","date_published":"2026-11-17T10:06:46.000+00:00","date_modified":"2026-11-17T10:20:38.589+00:00","author":{"type":"Organization","name":"IT之家","url":"https://www.ithome.com/1/028/453.htm"},"publisher":{"name":"Silicon Nexus","url":"https://silinex.xyz/"},"category":"chips","language":"zh","source_url":"https://www.ithome.com/1/028/453.htm","source_name":"IT之家","key_facts":["IT Home, November 17: European cloud service provider OVHcloud announced today local time the launch of its first family of AI models, Olala, and unveiled the first member of the series, Olala mini 7A1B.","Olala mini 7A1B uses a hybrid attention mechanism combining Differential-TPA and Mamba-3, with 7 billion total parameters and 1 billion activated parameters, in BF16 data format.","The model will be released as open source under the Apache 2.0 license on Hugging Face."],"key_facts_zh":["IT之家 11 月 17 日消息，欧洲云服务企业 OVHcloud 当地时间今日宣布推出其首个 AI 模型家族 Olala，同时公布了该系列的首个成员 Olala mini 7A1B。","Olala mini 7A1B 采用结合了 Differential-TPA 和 Mamba-3 的混合注意力机制，总参数 70 亿、激活参数 10 亿，使用 BF16 数据格式。","该模型将以 Apache 2.0 许可开源发布在 Hugging Face 上 。"],"tldr":["IT Home, November 17: European cloud service provider OVHcloud announced today local time the launch of its first family of AI models, Olala, and unveiled the first member of the series, Olala mini 7A1B.","Olala mini 7A1B uses a hybrid attention mechanism combining Differential-TPA and Mamba-3, with 7 billion total parameters and 1 billion activated parameters, in BF16 data format.","The model will be released as open source under the Apache 2.0 license on Hugging Face."],"tldr_zh":["IT之家 11 月 17 日消息，欧洲云服务企业 OVHcloud 当地时间今日宣布推出其首个 AI 模型家族 Olala，同时公布了该系列的首个成员 Olala mini 7A1B。","Olala mini 7A1B 采用结合了 Differential-TPA 和 Mamba-3 的混合注意力机制，总参数 70 亿、激活参数 10 亿，使用 BF16 数据格式。","该模型将以 Apache 2.0 许可开源发布在 Hugging Face 上 。"],"entities":[],"signal_type":"other","impact_score":0,"key_metrics":[],"body_md":"## Wire brief\n\nBased on public reports including IT之家: European cloud service provider OVHcloud announced today local time the launch of its first AI model family, Olala, along with the first member of the series, Olala mini 7A1B.\n\nBackground: Olala mini 7A1B uses a hybrid attention mechanism combining Differential-TPA and Mamba-3, with 7 billion total parameters and 1 billion activated parameters, using the BF16 data format. The model will be released as open source under the Apache 2.0 license on Hugging Face. It can run on GPUs with 8GB of VRAM, achieving token throughput of 160 TPS on an RTX A4500, and 50 TPS even on CPU-only platforms.\n\nPublic information also notes: OVH said Olala mini 7A1B was trained on three European public supercomputers using NVIDIA's GH200, A100, and H100, with total costs reaching 7 million euros (approximately 51.428 million yuan). The initial version of Olala mini 7A1B has a context length of 32K, and version 1.1, expected by the end of this year, will expand to 128K context and offer better mathematical capabilities. OVH will continue developing AI models, and Olala 2.0 will have a parameter scale of >100B.\n\nSilicon News Wire only rewrites and polishes key points; the complete wording and context are subject to the original page.\n\n---\n\nSource desk: **IT之家** · [origin link](https://www.ithome.com/1/028/453.htm)\n","body_md_zh":"## 电讯正文\n\n综合 IT之家 等公开报道整理：欧洲云服务企业 OVHcloud 当地时间今日宣布推出其首个 AI 模型家族 Olala，同时公布了该系列的首个成员 Olala mini 7A1B。\n\n事件脉络方面，Olala mini 7A1B 采用结合了 Differential-TPA 和 Mamba-3 的混合注意力机制，总参数 70 亿、激活参数 10 亿，使用 BF16 数据格式。该模型将以 Apache 2.0 许可开源发布在 Hugging Face 上 。其 可在 8GB 显存的 GPU 上运行 ，在 RTX A4500 上可实现 160 TPS 的词元吞吐量，仅 CPU 平台也能实现 50 TPS。\n\n公开信息还提到：OVH 表示，Olala mini 7A1B 在三台欧洲公共超级计算机上利用 NVIDIA 的 GH200、A100、H100 训练而成， 总开销达到 700 万欧元 （约合 5,142.8 万元人民币） 。Olala mini 7A1B 模型初始版本上下文长度为 32K，预计今年底推出的 1.1 版本将扩展至 128K 上下文并具备更好的数学能力。OVH 后续将持续开发 AI 模型， Olala 2.0 的参数规模将＞100B 。\n\n---\n\n来源台：**IT之家** · [原文链接](https://www.ithome.com/1/028/453.htm)\n","observations":{"human":24,"bot":10,"crawlers":[{"crawlerId":"Googlebot","hits":5,"lastSeenAt":"2026-11-18T16:26:48.652Z"},{"crawlerId":"AhrefsBot","hits":3,"lastSeenAt":"2026-11-18T07:55:48.497Z"},{"crawlerId":"ClaudeBot","hits":2,"lastSeenAt":"2026-11-20T09:57:32.387Z"}],"last_seen_at":"2026-11-20T09:57:32.387Z","note":"Counts are request logs classified by User-Agent. User-Agents can be spoofed. This is not proof of search-engine or LLM inclusion."}}