Home
Tools
Models
App
Prompts
News
Topics
Submit for inclusion
English
English 日本語 한국어 Português español Deutsch Русский Français 繁體中文 简体中文
option
Home Close
Home
Tools
Category Ranking Generate Cover Generate Image Skill
Models
Large Language Model Multimodal Model
App
Prompts
Prompts Image Prompts
News
News Flash News
Topics
Submit for inclusion
Submit App Submit Tool
Create an account Sign In
Change Language English Change Language
Settings
English EN 日本語 JA 한국어 KO Português PT español ES Deutsch DE Русский RU Français FR 繁體中文 ZH-TW 简体中文 ZH-CN
Home
Flash News
Content
KennethRodriguez
KennethRodriguez
July 8, 2026

SuperCLUE evaluated Tencent's Hy3 language model against DeepSeek-V4-Pro on coding tasks. Hy3 uses an MoE architecture with 295B total and 21B activated parameters, supporting 256K context. It scored 47.37, tying with DeepSeek-V4-Pro while being much smaller. Cost is just 0.43 yuan per question, speed under 400 seconds, and each task averages 40 conversation rounds and 1.16M tokens. Hy3 is fully open-sourced under Apache 2.0.

0
0
Share Click to copy the link
Comments (0)
0/300
Submit
Author PaulKing PaulKing September 28, 2026
Intellifusion launched ZCode compensation plan on Sep 28, offering quotas and tokens to restore trust after default code upload controversy. Shares dropped 2.05% in HK. The company confirmed repository snapshot path removal and third-party data deletion, ensuring local code storage. This shift highlights industry focus on data sovereignty and engineering security in AI developer tools.
0
0
Author EdwardSmith EdwardSmith September 28, 2026
Hackers are increasingly stealing AI models and computing power for ransomware and espionage. Google Threat Intelligence reports a surge in LLM hijacking, with dark web sellers offering premium AI subscriptions at up to 97% discount and guaranteed access. Criminals also hijack cloud servers to run free models. Threat actors from over 20 countries are misusing tools like Claude, making AI access a central underground economy. Experts warn that AI theft is rising rapidly and will become a critical component of both attack and defense strategies, urging immediate security preparedness.
0
0
Author DonaldBrown DonaldBrown September 28, 2026
OpenAI plans to launch permanent AI assistant O on Sep 29. Anthropic tests Claude Sonnet 5.5, rivaling GPT-6Astra with competitive pricing. MiniMax launches M3.1-Flash-Preview with developer incentives. Zhipu ZCode compensates users and removes code upload paths for security. Xiaomi MiMo-V2.6-Pro tops Code Arena rankings. Nubia NaviX Ultra faces game anti-cheat conflicts. Google tests direct Flipkart purchases via Gemini in India. NVIDIA releases free Nemotron3Diarization model for real-time voice segmentation.
0
0
/live/7657
Author GregoryRamirez GregoryRamirez September 28, 2026
NVIDIA released Nemotron3Diarization, a free lightweight AI model with 100 million parameters for speech segmentation. It tops VoiceArena benchmarks with 14.72% error rate and supports real-time processing of up to eight simultaneous speakers. The model offers dynamic audio buffer settings for latency-accuracy balance and integrates with Parakeet for automated transcription with anonymous speaker tags. This release lowers technical barriers for applications like real-time meeting records and smart customer service, accelerating multi-speaker identification deployment in edge devices and commercial scenarios.
0
0
/live/7656
Author BillyGarcía BillyGarcía September 28, 2026
Google tests in-app purchases on Flipkart via Gemini AI Mode in India, allowing users to buy electronics without leaving the interface. The limited rollout expands in late October to coincide with peak shopping season, supported by Google's $350 million stake in Flipkart. Currently restricted to select products and excluding Amazon, this move marks a shift from product discovery to direct transaction conversion, challenging e-commerce monetization models as the closed-loop ecosystem matures.
0
0
OR