瑞典初创公司 Strawberry 结束测试,正式推出内置 AI 助手的浏览器,定位为企业版 ChatGPT 和 Claude 的替代方案。用户可在浏览器内为销售、招聘、客服和营销等任务搭建名为 companion 的智能体,产品不绑定特定模型,智能体运行在用户设备上。订阅价格为每月 20 美元、85 美元和 250 美元三档,另有免费版;公司此前完成 600 万美元融资,投资方包括 General Catalyst 和 EQT Ventures。
一家获得 General Catalyst 以及 Lovable 和 Hugging Face 创始人投资的瑞典初创公司,今天将其内置 AI 助手的 AI 浏览器从测试阶段正式推出,并将自身定位为面向企业的 ChatGPT 和 Claude 替代方案。 AI 助手领域竞争激烈,投资者纷纷押注 AI 应用层初创公司,而这些公司由大语言模型背后的 AI 巨头提供支持。 总部位于斯德哥尔摩的 Strawberry 拥有 12 人的团队,希望将 AI 从开发者这一关键用例扩展到企业中的非技术团队。其理念是,通过自动化日常业务任务,让员工腾出时间从事更高价值的工作。 Strawberry 成立于 2023 年,将 AI 直接嵌入浏览过程,使没有编程知识的员工也能为销售、招聘、客户支持和营销等特定业务任务构建智能体,Strawberry 将其称为“companion”(伙伴)。 Strawberry 不绑定特定 AI 模型,还为企业提供现成的智能体,用户可以要求这些智能体执行业务任务。这些虚拟智能体会学习企业其他部分的运作方式,智能体的工作成果会在整个企业内共享,因此正如 Strawberry 所
e remaining compatible with torch.compile. Each frontier model expands TorchTitan’s reusable toolkit, from multimodal data pipelines, sample packing, shared vision components, and Hugging Face checkpoint interoperability to composable 5D parallelism, declarative sharding, optimized Attention Gym and Helion kernels, and rigorous numerical validation. Developers can build on this foundation to bring newly released architectures to validated, scalable training faster, entirely in native PyTorch. Ti
isms that expose the host kernel, and that it supports PrivateLink and serverless egress controls, backed by its serverless security investments. The announcement cited the OpenAI Hugging Face incident as underscoring the risks of managing GPU infrastructure separately from governed enterprise data. AIR usage and costs appear in Databricks billing system tables alongside other Databricks workloads, with workload-level usage records showing which training runs consume GPU budget. Robert Macy III,
cement, model, requirements). Persistent-agent services: LangChain’s Managed Deep Agents packages user-specific credentials, scoped memory, sandboxed execution and scheduled work; Hugging Face’s Agent Manager orchestrates persistent sessions across Claude Code, Codex and other harnesses (LangChain architecture, Agent Manager). Prompt caching: OpenAI added 12-hour API prompt caching, which can be mixed with the 30-minute default within a request. Longer retention carries extra cost but can benefi
cement, model, requirements). Persistent-agent services: LangChain’s Managed Deep Agents packages user-specific credentials, scoped memory, sandboxed execution and scheduled work; Hugging Face’s Agent Manager orchestrates persistent sessions across Claude Code, Codex and other harnesses (LangChain architecture, Agent Manager). Prompt caching: OpenAI added 12-hour API prompt caching, which can be mixed with the 30-minute default within a request. Longer retention carries extra cost but can benefi
rchTitan fine-tuning case, step time fell from 11.49s to 2.78s. DeferAndFuse alone gave 1.15x. Full torch.compile gave 3.80x. Model coverage: Google continuously tests about 5,000 Hugging Face models in eager and compiled modes. Fleet serving: Zenteiq grew daily output from 1.47B to 5.97B tokens per 1,000 TPU v6e chips. That gain came from llm-d routing on GKE, not TorchTPU alone. What integrations ship with TorchTPU? Hugging Face: Transformers v5 deploys to TPU with device map="tpu". Diffusers
llm-d 项目在 PyTorch Conference North America 上发布 1.0 版本,这是一个 Kubernetes 原生的分布式推理服务栈,由 Red Hat、IBM 和 Google 主导维护,CNCF 以沙箱项目托管。1.0 版本新增多租户服务质量保障,包括按优先级和公平性调度的流控与可控降载,并支持路由器热备、安全滚动更新,以及把 prefill 和 decode 副本作为整体管理的 DisaggregatedSet 控制器。项目公布的测试显示,在 544 张 NVIDIA H100 上部署 GLM-5.2 可服务最多 3000 个并发编码智能体且零抢占,按当前云租赁价格计算,每 token 成本比同等商业 API 低 5 到 10 倍。
nce stacks for any model, on any accelerator, in any cloud. Red Hat, IBM, and Google are the project’s lead maintainers and co-founded llm-d with CoreWeave and NVIDIA; AMD, Cisco, Hugging Face, Intel, Lambda, and Mistral AI have joined as contributors. The Cloud Native Computing Foundation (CNCF) hosts llm-d as a sandbox project. IBM describes the result as a vendor-neutral alternative for organizations that want what it calls frontier-class inference performance without depending entirely on on
IBM、Red Hat 与 Google 等 llm-d 项目贡献者发布分布式 AI 推理框架 1.0 版本,新增多租户服务质量、零停机升级和面向 AI 智能体的服务模式。框架基于 Kubernetes,可与 vLLM、SGLang 等推理引擎配合,通过路由把请求导向已缓存相关上下文的副本,并在 GPU 内存、CPU 内存和存储之间迁移缓存。IBM 称在一项智能体编码测试中路由器捕获了 97% 的可用缓存复用,IBM 与 Red Hat 还在自建基础设施上并发运行 3000 个软件工程智能体,成本比通过托管服务运行同一开源模型低 5 到 10 倍。
LC are its lead maintainers. They co-founded the project last year with CoreWeave Inc. and Nvidia Corp. Other contributors include Advanced Micro Devices Inc., Cisco Systems Inc., Hugging Face Inc., Intel Corp. and Mistral AI SAS. The project offers organizations seeking more control over inference costs and infrastructure choices a way to reduce dependence on a single model provider, cloud or hardware platform. “At IBM Research, we are seeing that open large language models and llm-d can greatl
esource allocation across stages vLLM-Omni is flexible and easy to use with: Heterogeneous pipeline abstraction to manage complex model workflows Seamless integration with popular Hugging Face models Tensor, pipeline, data and expert parallelism support for distributed inference Streaming outputs OpenAI-compatible API server Full-duplex realtime serving with streaming audio input and output vLLM-Omni seamlessly supports most popular open-source models on HuggingFace, including: Omni-modality mod
Google 发布 LiteRT Tensor API,让开发者在 C++ 中逐算子构建计算图,同一份代码可经 LiteRT 跑在 CPU、GPU(ML Drift)和 NPU 上,或不经模型文件直接通过 XNNPACK 在 CPU 运行。官方用该 API 手写 SAM 2 并部署到浏览器 WebGPU,在 MacBook Pro 上比 Apple MLX 快约 1.5 倍,输出掩码与 Meta PyTorch 参考逐像素一致。该 API 还支持从 .safetensors 直接加载权重,并让预处理、推理与后处理通过共享加速器缓冲区串联,避免数据回传 CPU。
or-optimized LLM composite ops, such as scaled dot-product attention, RMSNorm, and KV-cache updates. Direct weight ingestion: load weights straight from model checkpoints, such as Hugging Face .safetensors files, without conversion scripts needed in between. Tensor-centric buffer interop: pre-processing, model partitions, and post-processing pass tensors to each other through shared accelerator buffers, so intermediate data stays on the accelerator with zero copies. Building blocks for AI agents