iFlytek released its latest open-source model, Spark-X2.6, highlighting a fully autonomous stack built on domestic compute infrastructure from training to inference. The release aims to validate large-scale training feasibility outside the NVIDIA ecosystem, offering 1 billion free tokens via the Xingchen MaaS platform for developer testing.
On-Device Deployment
端侧部署
Running AI models locally on user devices instead of cloud servers.
Curated Intelligence
StepFun officially launched the STEPX Neo, a native AI agent smartphone designed to encapsulate cloud-based agents into a controlled hardware box via integrated software-hardware-model architecture. The device features retro-futuristic aesthetics and emphasizes deterministic physical interaction to address the instability of pure software agents.
Google DeepMind has released EmbeddingGemma 2, an open-source lightweight model built on the Gemma 4 architecture that maps text, images, audio, and video into a unified embedding space. Designed for consumer hardware, it enhances on-device search and multimodal RAG pipelines, building on the predecessor's success with over 20 million downloads.
Google DeepMind launched EmbeddingGemma 2, upgrading its previous text-only model to support unified embeddings for code, images, video, and audio. Built on the Gemma 4 architecture, it targets lightweight, privacy-first RAG pipelines on consumer hardware. The model is now open-sourced with guides for inference and fine-tuning.
Meta officially launched the Llama 3.2 series, introducing multimodal vision capabilities and optimized on-device performance for the first time. The release includes models ranging from 1B to 90B parameters, enabling developers to run vision-capable AI efficiently on local devices like smartphones and laptops.