Flash

The news, the context, and what changes.

RSS
Flash

Agnes AI Launches New High-Performance Text Model Agnes-2.5-Flash

Agnes AI has released Agnes-2.5-Flash, a new generation text model with coding capabilities ranked among the top globally. The launch coincides with the Agnes Code desktop application, which supports local project development. Tests indicate the model can quickly locate and fix hidden bugs, independently build complex web applications, and coordinate modifications across multiple files for large-scale projects.

88Z AI editorial1 min read
Flash

OpenAI Officially Launches Latest GPT-5.6 Large Language Model Series

OpenAI has officially released the GPT-5.6 series, featuring three models: Sol, Terra, and Luna, with API pricing ranging from $1 to $30 per million tokens. The flagship Sol model scored 53.6 in the Agents’ Last Exam benchmark, surpassing Claude Fable 5 by 13.1 points. It also achieved a record-high coding agent index of 80, while significantly reducing both cost and latency.

88Z AI editorial1 min read
Flash

Meta Opens Muse Spark AI Model API to Developers

Meta has officially opened the Muse Spark AI model API to developers and released the upgraded Muse Spark 1.1, marking its entry into commercial AI competition. As Meta's most powerful model for programming and agent tasks to date, it supports code generation, tool invocation, and multimodal understanding.

88Z AI editorial1 min read
Flash

OpenAI Launches ChatGPT Agent Workspace ChatGPT Work

OpenAI has released the new ChatGPT Work agent, powered by Codex and GPT-5.6. It supports cross-platform execution of multi-step, long-duration tasks, allowing users to hand over workflows via a single command to generate various documents and analyses.

88Z AI editorial1 min read
Flash

Ant Lingbo Tech Launches Industry's First Embodied-Native World Action Model LingBot-VA 2.0

Ant Lingbo Tech has released LingBot-VA 2.0, the industry's first embodied-native world action model. Built on a zero-based pre-trained autoregressive architecture, it integrates four core designs: semantic visual-action tokenizer, causal pre-training, MoE architecture, and asynchronous inference. It supports 150Hz real-time inference on a single card, enabling robots with general control capabilities to reason and act simultaneously.

88Z AI editorial1 min read
Flash

Kuaishou Launches Flagship Agentic Coding Model KAT-Coder-Pro V2.5

Kuaishou KwaiKAT has released its flagship Agentic Coding model, KAT-Coder-Pro V2.5, achieving breakthroughs in long-horizon engineering, general agentic capabilities, and large-scale reinforcement learning. By constructing over 100,000 runnable repositories via AutoBuilder and employing multi-framework RL with Mixture of Experts techniques, the model leads in benchmarks such as SWE-Bench Pro.

88Z AI editorial1 min read
Flash

ByteDance Launches Multimodal Image Creation Model Seedream 5.0 Pro

ByteDance's Seed team has released Seedream 5.0 Pro, significantly enhancing text-image alignment, text rendering, and aesthetic quality. The model achieves breakthroughs in visualizing complex information, interactive precise editing, realistic image texture restoration, and native multilingual generation. It supports converting data and dense text into professional layouts and allows for refined control via point-and-click selection.

88Z AI editorial1 min read
Flash

OpenAI Launches New Voice Model GPT-Live

OpenAI has released GPT-Live, a new voice model featuring real-time simultaneous interpretation, full-duplex conversation, and customizable reasoning intensity. It employs deep task delegation to handle complex background tasks while maintaining smooth interaction, and supports real-time visual card displays.

88Z AI editorial1 min read
Flash

SpaceXAI Launches New Flagship LLM Grok 4.5

SpaceXAI has released its flagship model, Grok 4.5, featuring a 1.5T parameter MoE architecture and joint training with Cursor. It rivals Claude Opus in performance, achieving 80 TPS inference speed and surpassing Opus 4.8 in programming on DeepSWE Bench, with over 4x better token efficiency. It currently supports a 500k context window, expanding to 1 million next week, with a 2T parameter version coming later this month.

88Z AI editorial1 min read
Flash

Ant Lingbo Open-Sources LingBot-Video, a Foundation Model for Embodied AI Video Generation

Ant Lingbo Technology has open-sourced LingBot-Video, the world's first Mixture-of-Experts (MoE) foundation model for embodied AI video generation. With 30 billion total parameters and only 3 billion activated during inference, it offers triple the efficiency of dense architectures. The model leverages a 70,000-hour dataset and multi-dimensional reinforcement learning rewards to improve physical plausibility and task completion.

88Z AI editorial1 min read
Flash

Mistral AI Launches Embodied AI Navigation Model Robostral Navigate

Mistral AI has released Robostral Navigate, an embodied AI navigation model that enables autonomous navigation in complex environments using only a single RGB camera. The 8-billion parameter model performs well in unknown environment tests and significantly reduces training time through technical optimizations, supporting continuous self-improvement.

88Z AI editorial1 min read
Flash

Claude Cowork Expands to Web and Mobile

Anthropic announced that the Claude Cowork task agent has officially expanded to web and mobile platforms, supporting cross-device session continuity and file operations. Users can now initiate tasks on desktop and monitor progress on mobile, with scheduled tasks capable of running offline in the background. Claude sends confirmation requests to users' phones for key decisions. The feature is currently in beta and initially available to Max subscribers.

88Z AI editorial1 min read