Google has made Gemini 4 Carbon generally available to developers, enterprises, scientists, and consumers, positioning it as its strongest frontier model for cybersecurity capabilities and long-horizon agentic workflows. Carbon sits above Gemini 4 Argon and targets extreme tasks such as complex multi-step coding, debugging, and open-ended scientific research, while Argon remains the default for most software engineering and knowledge work. Carbon was previously released to select cybersecurity defenders through the Fairwind Program.
Long-horizon Agent
长程智能体
A workflow paradigm where models autonomously plan, execute, and correct over long multi-step tasks.
Curated Intelligence
Tencent has officially released Hunyuan Hy4, a 770B-parameter model with 49B activated parameters that introduces native vision understanding and computer-operation capabilities for the first time, alongside upgraded long-context comprehension and long-horizon agent performance. It is already live across WorkBuddy, CodeBuddy, Yuanbao and WeCom, with API access via OpenRouter, and model weights slated for open release soon.
Tencent released Hunyuan Hy4, its latest flagship LLM, adding vision understanding for the first time and substantially scaling pretraining and post-training. First-day demand overwhelmed capacity, prompting a tenfold expansion of the WorkBuddy inference cluster and higher free quotas. Hy4 ranks second among Chinese models on Artificial Analysis' intelligence index and first among open models on Agent Arena, at roughly $0.78 per task.
Tencent unveiled Hunyuan Hy4, scaling its flagship LLM from under 300B to 770B parameters while adding vision understanding and Computer Use for the first time, with focused gains in instruction following, long-context comprehension and long-horizon agent tasks. Weights are slated for open release, pricing stays at RMB 6 per million input tokens and RMB 18 per million output tokens, and the model is already live across Tencent Cloud, Yuanbao, CodeBuddy and third-party platforms like OpenRouter.
Google DeepMind has introduced Gemini 4 Carbon, a frontier model designed for long-horizon agentic coding and cybersecurity. It ranks #1 on the Artificial Analysis Intelligence Index v5.0 and features a full loop for finding, validating, and patching vulnerabilities, currently available only to select security defenders.
Tencent's agent platform WorkBuddy has integrated Moonshot AI's newly released Kimi K3.1 Pro model, featuring 3 trillion parameters and a 1 million token context window. The integration also includes a high-speed variant offering 2-3x faster output, with significant enhancements in code generation, 3D creation, and long-horizon financial agent capabilities, achieving state-of-the-art results in multiple benchmarks.
Addressing high failure rates in enterprise AI pilots, Tezign founder Fan Ling proposes a dual world model architecture. The framework separates external customer preference modeling from internal process asset mapping to resolve agent deployment gaps, accompanied by the launch of a collaborative initiative.
vLLM Semantic Router introduces Session-Aware Agent Routing (SAAR), using router-held session memory and hard locks to resolve model-switching breaks in long-horizon LLM agents. Across 21,600 deterministic tests, it cut model switches by 79.29% and physical model costs by 78.71%, achieving zero session continuity violations on AMD ROCm hardware.