Anthropic Previews Mythos AI for Cyber Defense
Anthropic has debuted a preview of its powerful new AI model Mythos. The model launches within a cybersecurity initiative for defensive work by select high-profile companies.
Search
Tag: #ai-model12 results
Anthropic Tests Mythos Model with Apple, Amazon
Anthropic is granting access to its more powerful, unreleased Mythos AI model. Apple and Amazon are testing it to prepare for potential cyberattacks from wider AI availability.
WeChat Builds Own AI Model
Tencent's WeChat is developing a proprietary AI model, completing basics with 2026 launch planned for mini-programs and social features. Reduces third-party dependency amid competition from Deepseek and Doubao. Leverages 1.4B users and ecosystems for AI agents.
Xiaomi Chip-AI-OS Convergence Schedule Confirmed
Xiaomi confirms schedule for next-gen self-developed chip, AI large model, and OS converging in a terminal product, slightly later than rumors. Xuanjie O1 chip has shipped over 1 million units and will cover phone, tablet, car, wearables ecosystems. Lei Jun eyes 2026 for full self-chip/OS/AI integration in one device.
Alibaba Confirms HappyHorse AI Testing
Alibaba confirmed HappyHorse was developed by its Alibaba Token Hub (ATH) innovation unit and is in internal testing. API access will open soon as part of exploring new AI-era interactions.
Step 3.5 Flash 2603 Coding Model Launches
Jieyue Xingchen launches Step 3.5 Flash 2603, an optimized developer model for high-frequency programming and Agent workflows. It features specialized enhancements in code generation, debugging, refactoring, and Agent tasks. Step Plan subscribers can directly call the model API.
Experts Dismiss OpenClaw as Unexciting
AI experts question the hype around OpenClaw, calling it unremarkable. One researcher stated it offers nothing novel from an AI perspective. This tempers expectations post-launch.
Volcano Engine's Seed 2.0 Ambitions
Volcano Engine embeds its ambitions in Seed 2.0. The Doubao large model enters 2.0 phase, shifting to pragmatism. Focus moves from hype to practical utility.
Doubao 2.0 Cuts Inference Costs 10x
ByteDance released Doubao 2.0, a systematic upgrade for the Agent era. It matches GPT-5.2 and Gemini 3 Pro performance while reducing inference costs by an order of magnitude. This targets large-scale production for complex tasks.