Kling 4.0 is live on ArtArch 🎬 @Kling_ai Longer shots, finer control, smoother motion—let your creativity run wild! #Kling4 #KlingAI 30 creator spots are open: apply via @testskill001 with your work for exclusive Kling credits. Share your creation for a chance at bonus credits! https://t.co/qgl7NwlRqO
#Product updates#Video#Image generation
#English text is being prepared.#Speech#English text is being prepared.
HachimoDock, an open-source AI hardware incubated by the Innovative Product Team of Xiaomi's Technology Committee, made its debut during the 2026 Guancha Conference, with a DIY hardware cost of about 150 yuan. The device combines a desktop pet, voice control of PC-side Agents, smart home integration, and custom component creation, and can dispatch voice intents to Agents such as Codex and WorkBuddy while displaying task status in sync. The software source code is now public on GitHub, along with a hardware bill of materials, structural component models, and an assembly tutorial.
#Product updates#Open-source ecosystem#Agents
Jimeng AI, under ByteDance, has announced the official launch of the Jimeng Short Drama Studio, offering the in-house script model DreamScript, the image model Anycook, and the storyboard model DreamShot. DreamScript supports script generation across multiple genres, hit-potential diagnosis, and analysis of long scripts of up to 100,000 characters; Anycook is aimed at short drama art direction, tuning character designs and scenes; DreamShot supports both canvas and storyboard modes, as well as dynamic previews on a 3D director's console. The studio also offers post-production capabilities such as clip re-shoots, sound reconstruction, and intelligent shot cutting, plus a re-drawing feature that, after uploading original footage, re-creates the visuals according to style, aspect ratio, and target region.
#Product updates#Video#Multimodal
Agibot has launched the safety-standard heavy-load embodied robot Genie G2 Max, with a single-arm payload of 20kg, a dual-arm rated payload of 40kg, and a peak of up to 50kg. Its arm repeat positioning accuracy is ≤±0.5mm, and it targets scenarios such as depalletizing and palletizing, material transfer, and production line feeding. The product is designed to be compatible with existing production lines, with a vertical working range of 0.1m to 2.5m. It supports single-side depalletizing and palletizing and pallet placement against walls, and features dual-battery hot swapping and autonomous recharging to support 7×24-hour operation. On the Qiyuan T1 production line, the G2 Max is responsible for transporting joint components, matching the takt time of one unit rolling off the line every 2.5 minutes, and natively adapts to the Genie Studio 2.0 embodied operation platform.
#Product updates#Embodied AI#English text is being prepared.
Cloudflare has opened a public beta of Auto Router in AI Gateway; after setting the model to cloudflare/auto, a classifier selects a model for each request. Routing has two stages: first, a candidate pool is filtered by request format, credentials, billing, and quota; then a multi-head classification model on Workers AI scores the request across 14 task categories and rates it from 1 to 5 on four dimensions—complexity, ambiguity, risk, and dependence on prior context—combining benchmark results according to "utility = expected quality - adaptive cost penalty." Classification reads only the most recent few turns, capped at 8 KiB of text; responses return cf-aig-routed-model and cf-aig-routing-reason for post-hoc auditing, and request headers can also be used to restrict candidate models. The beta is free and does not yet support WebSockets.
#Product updates#English text is being prepared.#OpenAI
Cadence announced a partnership with Nvidia to launch the Maestro AI model family, fine-tuned on Nvidia's Nemotron series of models, for understanding and invoking Cadence's own EDA tools. Cadence says that in early testing, Maestro improved output quality by up to 30% and token efficiency by up to 5x compared with general-purpose foundation models. The Maestro platform also supports customization based on customer data, and the models and agents can be invoked via API, run in a secure VPC, or be deployed on-premises.
#Model releases#Agents#English text is being prepared.
Swedish startup Strawberry has exited beta and officially launched a browser with a built-in AI assistant, positioned as an alternative to enterprise versions of ChatGPT and Claude. Users can build agents called companions within the browser for tasks such as sales, recruiting, customer service and marketing; the product is not tied to a specific model, and the agents run on the user's device. Subscription prices come in three tiers at $20, $85 and $250 per month, with a free version also available; the company previously raised $6 million in funding from investors including General Catalyst and EQT Ventures.
#Product updates#Agents#English text is being prepared.
On October 26, Jimeng AI launched "Short Drama Studio," integrating script organization, character and scene management, storyboard generation, video production, and post-production adjustments into a single platform. The feature is powered by DreamScript, Jimeng's first professional script model, and supports uploading long texts of up to 100,000 characters and continuous generation of 100 episodes; the Script Doctor can automatically diagnose the main storyline of the first three episodes. Some features integrate the same capabilities as the Xiaoyunque short drama Agent, and a team edition supporting multi-account assets and unified credit management has also been launched.
#Product updates#Video#Agents
At the 2026 Global 1024 Developer Festival, iFlytek released the device-cloud integrated Spark X2.6, expanded the context window to one million Tokens, and open-sourced the Spark X2.6-4B on-device model, with device-cloud routing accuracy increasing from 81.2% to 96.1%. At the same time, it debuted the Agent OS agent collaboration platform and the Xingchen TaaS platform; the former is said to improve cross-system collaboration efficiency by 90% and enable more than 75% of processes to flow autonomously, while the latter builds a full-chain system from model capabilities to industry solutions. The iFlytek Open Platform also launched the XFYUN CLI development workbench, where a single command can invoke more than 700 AI atomic capabilities and more than 200 agent adaptation capabilities.
#Product updates#Agents#Speech
AgiBot has released the safety-standard heavy-load embodied robot Genie G2 Max, featuring dual arms with a rated payload of 40 kg and repeat positioning accuracy at the sub-millimeter level. The robot adopts 7-DOF force-controlled dual arms and can lean forward against walls for operation via a waist pitch joint, with a vertical operating range covering 0.1 meters to 2.5 meters. In e-commerce warehousing, it can complete 80 boxes of depalletizing and 75 boxes of palletizing per hour. The G2 Max natively supports Genie Studio 2.0, the embodied intelligence operation platform released by AgiBot last week, and has already taken on the task of transporting heavy joint modules on AgiBot's Qiyuan T1 joint assembly line.
#Product updates#Embodied AI#Agents
During AI WAVE, Zhiyuan Robotics released the safety-compliant heavy-load embodied robot Genie G2 Max, featuring a single-arm payload of 20 kg, a dual-arm peak payload of 50 kg, and arm repeat positioning accuracy of ≤±0.5 mm. The robot has been validated and deployed on production lines of daily chemical FMCG, e-commerce, and food companies, covering processes such as depalletizing and palletizing, case opening and loading, empty case recovery, and warehouse inbound; on the Qiyuan T1 production line of Huguang Co., Ltd., it matches the takt time of one robot rolling off the line every 2.5 minutes.
#Product updates#Embodied AI
iFlytek Spark-X2.6 has launched on the iFlytek Xingchen MaaS platform. It uses a MoE architecture with 293B total parameters and approximately 30B activated per inference, supports a 1M context, and is priced at a 40% discount via API at 0.9 yuan per million input tokens and 3.6 yuan for output. Compared with X2.5 from September, the company claims code and agent capabilities each improved by about 8 points, domestic training platform efficiency increased 4x, and the four-layer KV Cache hit rate reached 95%. The author tested the newly open-sourced Spark-X2.6-4B running an business review task offline via Ollama on a computer without a discrete GPU, taking 6.8 minutes; the author then used the local 4B to desensitize a contract before sending it to the cloud for review, consuming 7,368 tokens. A quota of 1 billion tokens is currently available to claim on the Xingchen MaaS platform.
#Product updates#Model releases#Agents
WeCom and WorkBuddy are now officially integrated, with AI directly entering the WeCom sidebar and contacts. WeCom accounts support one-click login, with no need to switch to external software or download an additional client. Within the authorized scope, WorkBuddy can read and invoke messages, files, calendars, and emails in WeCom, and generated content and documents can flow directly back into WeCom conversations, WeDrive, and business workflows. The article also presents implementation scenarios such as customer lead aggregation and customer locking, brokerage report generation, and mold cost pricing for manufacturing enterprises.
#Product updates#Agents#English text is being prepared.
Researchers from Tsinghua University and the Chinese University of Hong Kong have open-sourced AKV, a KV cache engineering service system built on SGLang that lets agents directly discard selected KV cache entries and reuse the rest, rather than summarizing old history. Across 3 models and 4 agent benchmarks, average cost dropped by 27.7% and average task scores improved by 1.7 points. The code is released under the Apache-2.0 license and currently supports only the Qwen, GPT-OSS, and MiniMax-M2 text models.
#English text is being prepared.#Agents#English text is being prepared.
HachimoDock, an open-source AI hardware incubated by Xiaomi's Technology Committee Innovative Products Team, officially debuted at the 2026 Guancha Conference in Hangzhou, with a hardware DIY cost of about 150 yuan. The device integrates an AI desktop pet, voice control of PC-side Agents, smart home integration, and custom component creation. Through voice, it can dispatch tasks such as looking up information, organizing files, and writing code to Agents including Codex and WorkBuddy, while the screen displays execution status in sync. The software source code is now public on GitHub, along with a hardware bill of materials, structural component models, and assembly tutorials, supporting self-replication and secondary development.
#Product updates#Open-source ecosystem#Agents
HachimoDock, an open-source AI hardware incubated by the innovative product team of Xiaomi's Technology Committee, made its debut at the 2026 Guanzha Conference, with a DIY hardware cost of about 150 yuan. The device integrates a personalized desktop pet, voice control of PC-side Agents, smart home integration, and custom component creation. Users can collaborate with a Coding Agent in natural language to generate components such as to-do lists, computer status dashboards, and Pomodoro timers, and can also dispatch requests via voice—such as looking up information, organizing files, and writing code—to Agents including Codex and WorkBuddy. The device can connect with Xiaomi MiloCo for voice-based home control. The software source code is now public on GitHub, along with a hardware bill of materials, structural component models, and an assembly tutorial.
#English text is being prepared.#Agents#Embodied AI
@Kling_ai's Kling 4.0 is live on Runware on D0 🎬 Up to 30 seconds of 1080p video with native audio, plus: - Multi-keyframe support, including first, middle, and last frames - Up to 15 inputs per request across keyframes, reference images, reference videos, and elements - Prompt-based editing of existing clips From $0.133 per second of output, with 15% off regular pricing for the next 7 days. It joins the 19 Kling models already on our API, across video, image, avatar, and lip-sync. Already calling Kling models on Runware? Switch your model ID to klingai:kling-video@4.0
#Product updates#Video#Tutorials / practice
HachimoDock, an open-source AI hardware incubated by the Innovative Products team of Xiaomi's Technology Committee, debuted during the 2026 Guancha Conference, with a DIY hardware cost of about 150 yuan. The device combines desktop pet interaction, voice control of PC-side Agents such as Codex and WorkBuddy, linkage with Xiaomi Miloco smart home devices, and on-site generation and running of custom components; task status is fed back through the screen and the desktop pet. The software source code is now public on GitHub, along with a hardware bill of materials, structural component models and assembly tutorials.
#Product updates#Open-source ecosystem#Agents
Open weights. Shared progress. MiniMax H3 is moving fast. We built H3 for video generation with native stereo audio and multimodal reference control. The open-source community is making that capability faster, more accessible, and easier to build on. Recent highlights: • FastH3 — FastVideo, Nuva Lab and NVIDIA: 4-step distillation, now running on DGX Spark and Apple Silicon. • Sol-H3 — NVIDIA’s SANA team: 15 seconds of 768p video + audio in 6.6 seconds on 8×B300, in the team’s warm-inference benchmark.* • VDN — Haocheng Xi and the OpenVDN team: rethinking attention for faster H3 inference, with weights, training and inference code released. • PDD — NVIDIA’s distillation method, brought to H3 by Alibaba PAI as 8-step Acc-LoRAs, now supported in ComfyUI. • LightX2V — 4- and 8-step Turbo LoRAs, with workflows for text, image and reference-conditioned video + audio. Behind every release are people training, optimizing, quantizing, testing and sharing. Special thanks to: @haoailab @nuvalab @NVIDIAAI @xieenze_jr @HaochengXiUCB @ArashVahdat @julberner @LightX2V @ComfyUI And to the individual contributors pushing the work forward: @haozhangml @cxlcl1 @lawrence_cjs @yitongli165665 @haopengl33 @songhan_mit @shanasaimoe Thank you for building with H3 and helping make it faster, more accessible, and more useful for the community. Powerful models go further when we build together. Keep pushing H3. Excited to see what comes next. 🚀 Explore the ecosystem: https://t.co/F1S5IjUNi1
#English text is being prepared.#Video#Open-source ecosystem