AI DayaHimour
AR
NEWS
China Blocks Meta's Acquisition of AI Startup Manus— Meta DeepSeek Launches Open-Source V4 Pro and Flash with Million-Token Context— DeepSeek OpenAI Launches GPT-5.5 with Advanced Coding and Agentic Research— OpenAI xAI Launches Grok Voice Think Fast 1.0, Scores 67.3% on τ-voice Bench— xAI MiMo-V2.5-Pro Launches: Xiaomi's Most Capable Agent Model Uses 40–60% Fewer Tokens Than Opus 4.6— Xiaomi OpenAI Launches ChatGPT Images 2.0 with Advanced Reasoning Mode— OpenAI Moonshot AI Launches Kimi K2.6: Open-Source Model Outperforms GPT-5.4 and Claude Opus 4.6 on Long-Horizon Coding and Agents— Moonshot AI Anthropic Launches Claude Opus 4.7: Agentic Coding Leap with 3.75MP Vision— Anthropic China Blocks Meta's Acquisition of AI Startup Manus— Meta DeepSeek Launches Open-Source V4 Pro and Flash with Million-Token Context— DeepSeek OpenAI Launches GPT-5.5 with Advanced Coding and Agentic Research— OpenAI xAI Launches Grok Voice Think Fast 1.0, Scores 67.3% on τ-voice Bench— xAI MiMo-V2.5-Pro Launches: Xiaomi's Most Capable Agent Model Uses 40–60% Fewer Tokens Than Opus 4.6— Xiaomi OpenAI Launches ChatGPT Images 2.0 with Advanced Reasoning Mode— OpenAI Moonshot AI Launches Kimi K2.6: Open-Source Model Outperforms GPT-5.4 and Claude Opus 4.6 on Long-Horizon Coding and Agents— Moonshot AI Anthropic Launches Claude Opus 4.7: Agentic Coding Leap with 3.75MP Vision— Anthropic
Latest News

Latest News

Follow the latest breaking news and developments in the world of artificial intelligence in real-time

Featured News

ArticleDeepSeek
Apr 24

DeepSeek Launches Open-Source V4 Pro and Flash with Million-Token Context

DeepSeek released two models: V4-Pro with 1.6T parameters (49B activated) and V4-Flash with 284B parameters (13B activated), both featuring a hybrid attention architecture that reduces compute per token to 27% of the previous generation. In maximum thinking mode, Pro-Max surpassed GPT-5.4 with a Codeforces Elo of 3206 and topped LiveCodeBench at 93.5%. Both models are available as open-source previews on Hugging Face.

DeepSeek
ArticleOpenAI
Apr 23

OpenAI Launches GPT-5.5 with Advanced Coding and Agentic Research

OpenAI released GPT-5.5 just two months after GPT-5.4, focusing on improved coding, computer use, and autonomous research capabilities. The model is available to paid subscribers via ChatGPT and Codex, classified as high-risk after extensive safety testing.

OpenAI
ArticlexAI
Apr 23

xAI Launches Grok Voice Think Fast 1.0, Scores 67.3% on τ-voice Bench

The model scored 67.3% on τ-voice Bench, over 20 points ahead of Gemini 3.1 Flash Live. It powers Starlink support with 70% self-resolution and 20% sales conversion.

xAI Blog
ArticleXiaomi
Apr 22

MiMo-V2.5-Pro Launches: Xiaomi's Most Capable Agent Model Uses 40–60% Fewer Tokens Than Opus 4.6

Xiaomi released MiMo-V2.5-Pro into public beta today, its most advanced model yet for long-horizon agentic tasks. The model sustains over 1,000 consecutive tool calls without context drift, completed a full SysY compiler in Rust in 4.3 hours with a perfect 233/233 test score, and built an 8,192-line desktop video editor across 11.5 hours of autonomous work. Benchmarks include 57.2% on SWE-bench Pro and 68.4% on Terminal-Bench 2.0. On ClawEval, it reaches 63.8% pass^3 using roughly 70K tokens per trajectory — 40–60% fewer than Claude Opus 4.6, Gemini 3.1 Pro, and GPT-5.4. Pricing remains unchanged at $1 per million input tokens and $3 per million output tokens. Open-source release is planned soon.

Xiaomi MiMo
ArticleOpenAI
Apr 21

OpenAI Launches ChatGPT Images 2.0 with Advanced Reasoning Mode

OpenAI announced the new image generation model ChatGPT Images 2.0 (gpt-image-2), which combines high visual fidelity with intelligent reasoning capabilities such as web search, generating multiple options, and self-reviewing results. The model is now available to all ChatGPT and Codex users and via API, supporting up to 2K resolution with better control over text and complex designs. 'Thinking' mode is available for Plus, Pro, and Enterprise users.

OpenAI Blog
ArticleMoonshot AI
Apr 20

Moonshot AI Launches Kimi K2.6: Open-Source Model Outperforms GPT-5.4 and Claude Opus 4.6 on Long-Horizon Coding and Agents

Moonshot AI has released Kimi K2.6 (1T MoE, 32B active) supporting a 262K-token context window and parallel agent swarms of up to 300 sub-agents. It scored 54.0% on Humanity's Last Exam (with tools), 83.2% on BrowseComp, 58.6% on SWE-Bench Pro, and 86.3% on BrowseComp with Agent Swarm. The model is now available on kimi.com, OpenRouter, Hugging Face, and Ollama.

AI DayaHimour
ArticleAnthropic
Apr 16

Anthropic Launches Claude Opus 4.7: Agentic Coding Leap with 3.75MP Vision

Anthropic officially launched Claude Opus 4.7 on April 16, 2026. The new model hits 64.3% on SWE-Bench Pro beating GPT-5.4, features 3.75MP high-resolution vision, and introduces the new xhigh effort level. Available now via Claude.ai, API, and Claude Code at $5 per million input tokens.

AI DayaHimour
8 news items

All News

ArticleMeta
Apr 27

China Blocks Meta's Acquisition of AI Startup Manus

China's National Development and Reform Commission issued a final decision blocking the $20 billion acquisition, citing national-security concerns and the prevention of strategic technology leakage. The ruling forced Meta to restructure its investments and unwind the Manus team integration.

National Development and Reform Commission (NDRC)
ArticleDeepSeek
Apr 24

DeepSeek Launches Open-Source V4 Pro and Flash with Million-Token Context

DeepSeek released two models: V4-Pro with 1.6T parameters (49B activated) and V4-Flash with 284B parameters (13B activated), both featuring a hybrid attention architecture that reduces compute per token to 27% of the previous generation. In maximum thinking mode, Pro-Max surpassed GPT-5.4 with a Codeforces Elo of 3206 and topped LiveCodeBench at 93.5%. Both models are available as open-source previews on Hugging Face.

DeepSeek
ArticleOpenAI
Apr 23

OpenAI Launches GPT-5.5 with Advanced Coding and Agentic Research

OpenAI released GPT-5.5 just two months after GPT-5.4, focusing on improved coding, computer use, and autonomous research capabilities. The model is available to paid subscribers via ChatGPT and Codex, classified as high-risk after extensive safety testing.

OpenAI
ArticlexAI
Apr 23

xAI Launches Grok Voice Think Fast 1.0, Scores 67.3% on τ-voice Bench

The model scored 67.3% on τ-voice Bench, over 20 points ahead of Gemini 3.1 Flash Live. It powers Starlink support with 70% self-resolution and 20% sales conversion.

xAI Blog
ArticleXiaomi
Apr 22

MiMo-V2.5-Pro Launches: Xiaomi's Most Capable Agent Model Uses 40–60% Fewer Tokens Than Opus 4.6

Xiaomi released MiMo-V2.5-Pro into public beta today, its most advanced model yet for long-horizon agentic tasks. The model sustains over 1,000 consecutive tool calls without context drift, completed a full SysY compiler in Rust in 4.3 hours with a perfect 233/233 test score, and built an 8,192-line desktop video editor across 11.5 hours of autonomous work. Benchmarks include 57.2% on SWE-bench Pro and 68.4% on Terminal-Bench 2.0. On ClawEval, it reaches 63.8% pass^3 using roughly 70K tokens per trajectory — 40–60% fewer than Claude Opus 4.6, Gemini 3.1 Pro, and GPT-5.4. Pricing remains unchanged at $1 per million input tokens and $3 per million output tokens. Open-source release is planned soon.

Xiaomi MiMo
ArticleOpenAI
Apr 21

OpenAI Launches ChatGPT Images 2.0 with Advanced Reasoning Mode

OpenAI announced the new image generation model ChatGPT Images 2.0 (gpt-image-2), which combines high visual fidelity with intelligent reasoning capabilities such as web search, generating multiple options, and self-reviewing results. The model is now available to all ChatGPT and Codex users and via API, supporting up to 2K resolution with better control over text and complex designs. 'Thinking' mode is available for Plus, Pro, and Enterprise users.

OpenAI Blog
ArticleMoonshot AI
Apr 20

Moonshot AI Launches Kimi K2.6: Open-Source Model Outperforms GPT-5.4 and Claude Opus 4.6 on Long-Horizon Coding and Agents

Moonshot AI has released Kimi K2.6 (1T MoE, 32B active) supporting a 262K-token context window and parallel agent swarms of up to 300 sub-agents. It scored 54.0% on Humanity's Last Exam (with tools), 83.2% on BrowseComp, 58.6% on SWE-Bench Pro, and 86.3% on BrowseComp with Agent Swarm. The model is now available on kimi.com, OpenRouter, Hugging Face, and Ollama.

AI DayaHimour
ArticleAnthropic
Apr 16

Anthropic Launches Claude Opus 4.7: Agentic Coding Leap with 3.75MP Vision

Anthropic officially launched Claude Opus 4.7 on April 16, 2026. The new model hits 64.3% on SWE-Bench Pro beating GPT-5.4, features 3.75MP high-resolution vision, and introduces the new xhigh effort level. Available now via Claude.ai, API, and Claude Code at $5 per million input tokens.

AI DayaHimour