NEWS & INSIGHTS

Updates & model insights

Follow platform announcements, model launches, and capability updates from official sources.

Kling AI Kling 4.0 Officially Launches, Focusing on High Dynamics and Complex Camera Movements

Kling AI's next-generation video generation model Kling 4.0 has officially launched, with realism, controllability, and professionalism as its core directions, focusing on high-dynamic and complex camera-movement scenarios. The new version improves continuity in high-speed motion and shot choreography, supports native generation of up to 30 seconds per run, and enables long takes and continuous storytelling; Omni Reference can combine up to 15 assets and supports up to 10 keyframes. The model also introduces output capabilities such as 10-bit HDR, 4K, and 21:9 aspect ratios, and integrates MCP and CLI, advancing AI video toward professional expression tools.

2026-10-16Read more

LATEST UPDATES

Latest updates

Platform Announcements · 可灵AI

Kling AI Kling 4.0 Officially Launches, Focusing on High Dynamics and Complex Camera Movements

Platform Announcements · Google

Google Gemini 4 Argon Global Launch, Promotional API Prices as Low as One-Fifth of Competitors

Google announced on October 13 (Eastern Time) that its flagship model, Gemini 4 Argon, is officially open to global users. Previously, the model was only provided selectively to cybersecurity partners. Ordinary users need to subscribe to Google AI Pro or Ultra to experience it, while developers can access it through a paid API. The new model's maximum output limit has been increased to 1 million tokens, with key optimizations in programming, complex reasoning, and robot control capabilities. During the promotional period, API pricing is highly competitive, with input and output prices at only one-fifth of major competitors, aiming to accelerate the adoption of enterprise workflows and research scenarios.

· 谷歌

Google Releases Gemini 3.8 Live and Extended Thinking Real-Time Conversation Models

On September 15 local time, Google announced the launch of two real-time conversation models: Gemini 3.8 Live and Gemini 3.8 Live Extended Thinking. The former balances intelligence with cost efficiency, while the latter enhances multi-step reasoning capabilities. Both models lead in voice quality and benchmark tests, support near-real-time visual processing and switching between 97 languages, and integrate SynthID watermarking technology. They are now available to developers and users via API, Studio, and various subscription services.

· OpenAI

OpenAI Releases GPT-Live-1 API: Supports Full-Duplex Real-Time Voice Interaction

OpenAI announced the introduction of GPT-Live-1 to its API, a model that integrates voice understanding and output, supporting full-duplex conversations and phone scenarios. It significantly reduces latency compared to traditional architectures and improves handling of interruptions and noise tolerance. The frontend voice layer is priced at $0.05 per minute, with new voice options added, aiming to simplify the process for developers building real-time voice agents.

· OpenAI

OpenAI Releases GPT-6 Astra: First Model to Reach Critical Cybersecurity Threshold

On September 3 local time, OpenAI announced the launch of its next-generation large language model, GPT-6 Astra. This model is the first under OpenAI to reach the "Critical" level cybersecurity capability threshold in its Preparedness Framework. Greg Brockman stated that with the release of Astra, the world has entered a new era of Artificial General Intelligence (AGI). Compared to its predecessor, GPT-5.6 Sol, Astra shows significant improvements in vulnerability discovery, jailbreak resistance, and alignment safety. Although observability has decreased, the overall likelihood of violating safety constraints is lower.

· Anthropic

Anthropic Releases Claude Fable 5.1 and Mythos 5.1: Performance Surpasses Previous Generations

On September 1 local time, Anthropic announced the launch of the Claude Fable 5.1 and Claude Mythos 5.1 models, positioned as the world's most advanced models for programming and knowledge work. Fable 5.1 is open to all users, performing excellently in benchmarks such as Terminal-Bench, and has reduced cache read fees. Mythos 5.1 is available exclusively through a trusted access program to cybersecurity and life science institutions, demonstrating breakthroughs in fields such as molecular design and computational analysis. Additionally, Anthropic launched enterprise-grade frontier safeguards (EFS) and implemented watermarking requirements under the EU AI Act.

· 阿里研究院

Qwen3.8-Max Upgraded to Version 0902, Enhancing Programming and Office Capabilities

Alibaba Research announced the upgrade of Qwen3.8-Max to version 0902. Building on prior open-source feedback, this version underwent further post-training tailored for programming and professional office scenarios, significantly boosting overall performance and better aligning with enterprise complex tasks, scientific research, and long-cycle workflows. The new version claimed first place on the CodeArena Frontend Programming Leaderboard, setting new records in reasoning, tool calling, and end-to-end application generation capabilities, while reducing the comprehensive cost per million tokens to $5. The model is now live on the Qwen AI Platform offering API services, and has been simultaneously integrated into Qwen Office, Qoder, and the Qwen APP.

· MiniMax

Minimax H3 Max Video Model Now Live

The AGTCloud platform has officially integrated the Minimax H3 Max video generation model, supporting video content creation at speeds faster than real-time. Users can generate videos via text descriptions or by uploading source images. The system features high prompt adherence, accurately restoring subject characteristics, motion trajectories, and camera directions. It supports custom start and end frames to control scene progression, automatically outputting rigorously composed, well-lit, and cinematic-quality visuals to meet diverse visual production needs.

· Z.ai

GLM-5.3 Announces Safety Evaluation Progress, Open Weights Release Delayed

Z.ai announces progress on the safety evaluation of the new generation GLM-5.3; due to the model's improved capabilities in cybersecurity tasks, the release of open weights will be delayed to complete further testing and protection.

· DeepSeek

DeepSeek V4 Pro Preview Release: Million-Token Context and Enhanced Agent Capabilities

DeepSeek V4 Pro is designed for complex reasoning, mathematics, coding, and agent tasks, with the official API offering a 1M context window and both thinking and non-thinking modes.

· ByteDance

Seedream 5.0 Lite: From Image Generator to Visual Creation Assistant

Seedream 5.0 Lite strengthens understanding, reasoning, generation, and editing through a unified multimodal architecture, and adds web search enhancement.

· 谷歌

Google to Release Gemini 3.8 Flash Model with Upgraded Coding Capabilities

According to The Wall Street Journal, Google is set to release a new model codenamed "Skimaki," Gemini 3.8 Flash, expected to go live on Wednesday local time. In internal tests, the model's coding performance surpassed Anthropic's Opus, significantly narrowing the gap with flagship models from OpenAI and Anthropic. This release comes amid high-level personnel changes at DeepMind and the departure of several core researchers, as Google accelerates its catch-up efforts through reinforcement learning and talent recruitment.

· OpenAI

OpenAI Releases ChatGPT Images 2.5: 50% Latency Reduction and New Sketch Feature

On September 9, OpenAI officially announced the release of its current state-of-the-art image generation model, ChatGPT Images 2.5. This version improves detail sharpness, lighting, and texture while reducing generation latency by up to 50%. New features include Sketch reference, a template library, and in-image comment annotations, supporting more precise multi-turn editing control. The API offers two models, Flare and Sunburst, suitable for rapid generation and high-demand workflows respectively, and continues the C2PA metadata watermarking security mechanism.