AI News

Google’s Gemini 3.8 Flash improves agent and software engineering performance while retaining extremely fast output. It scores 59 on Artificial Analysis’s Intelligence Index, ties several larger frontier models, and reaches 74% on DeepSWE v1.1. More reasoning and tool calls raise its average benchmark task cost by about 40% to $0.58, but the model remains highly cost-efficient for its capability level. Google also introduced Gemini 3.8 Flash Cyber for vulnerability discovery and patching, with restricted initial access.

A report says OpenAI’s alleged GPT-6, internally called Astra, may use recurrent-depth reasoning that repeatedly processes representations in the same network block. The approach could improve coding and computer-use performance while reducing parameter, memory and bandwidth requirements. Related research suggests small models can use additional test-time computation to approach the compute of much larger models. The article also highlights a safety trade-off: reasoning carried out in latent space may be less readable and therefore harder to monitor.

This AI Daily roundup covers Jimeng Studio's long-form AI film support, DeepSeek's reported RMB 475 million in seven-month revenue, Manus data recovery, Claude Cowork memory interoperability, OpenAI's reported Jalapeno chip, Stability AI funding, Grok Bot updates, and OpenRouter's asynchronous video API.

AI is moving into physical environments as robotics models become more reusable and hardware becomes cheaper. Yunzheng Li argues that long-term value will shift from robot bodies to intelligent models, real-world data, middleware, tools, and developer ecosystems. He also describes Agent-native development, where AI agents configure environments and run robotics demos. China’s manufacturing strengths and the United States’ model and software ecosystems may provide complementary foundations for global robotics platforms.

Hugging Face lists DeepSeek-V4-Flash-Vision-Exp as part of the nine-item DeepSeek V4 collection. The page was updated approximately 17 hours before capture and displays a count of 864. The supplied text does not include technical details such as architecture, benchmarks, licensing, weights, or deployment guidance.

Tencent expanded the Hy4preview inference cluster after the model’s WorkBuddy debut triggered heavy demand and queueing. The company also extended limited-time free access to Hy3 and other existing models through September 30, 2026, while warning that peak-time queues may recur.

DeepSeek may increase V4 Flash prices to reduce overload rather than cover operating losses. Third-party providers can undercut the official API by running the model’s open weights themselves, but coding tests suggest OpenCode Go may consume two to ten times more value than direct API access because of weaker cache reuse. DeepSeek’s prefix persistence, common-prefix detection, Token-wise Compression, Sparse Attention and compressed KV Cache contribute to lower long-context costs. The article also highlights the importance of the coding Harness, including routing, memory, context compaction and tool execution, in determining how much of the model’s capability users actually receive.

Hugging Face lists allenai/MolmoAct2-SO100_101 as a 5B robotics model. The page was updated on May 23 and records about 5.87k uses and a displayed value of 22. The excerpt includes no technical or evaluation details.

This review examines Transformer-based large models and self-supervised learning for battery prognostics and health management. It covers data scarcity, cross-domain generalization, interpretability, robustness, automation, and the challenges of trustworthy industrial and on-device deployment.