China AI

AIvLLM says it delivered 7.8x or more the throughput of GB200 on SemiAnalysis AgentX, serving MiniMax M3 on Vera Rubin NVL72 at matched interactivity.

AIvLLM says users can try it on Rubin today with the vllm/vllm-openai:cu134-nightly image, which runs DeepSeek, Kimi, GLM, and MiniMax models. The post links to a blog preview and says further improvements for Vera Rubin are planned.

AISara Hooker, who owns the account linked to Cohere, posts "Intriguing" and says she wants to be a Chinese VIP for a day. The post gives no details about what the VIP experience involves or what prompted the remark.

AIStepFun will host a Step 5 Preview Launch Night in San Francisco on October 22, where attendees can meet the team and see the Step 5 Preview in action. Registration is available through a Luma event page.

AIKurt Campbell, a former White House Indo-Pacific coordinator and US deputy secretary of state, is the subject of this interview on US China policy, the Indo-Pacific Quad and the risks of AI. The source text provided is truncated after his background, so it does not show his specific views on these topics.

AITinker has cut prices up to 70% on long-context prefill and sampling, which now cost the same as short context.

AISGLang says it worked with NVIDIA to optimize attention, MoE, and speculative verification kernels for Kimi K3 inference on early-access Rubin hardware.

AIHiDream-O1-Video-1.0 ranks #6 in Artificial Analysis's Image to Video with Audio leaderboard, just behind Dreamina Seedance 2.0 720p.

AIArtificial Analysis invites users to check HiDream-O1-Video-1.0 on its AA-Video-I2V v1.0 image-to-video leaderboard. Users can also vote for the model in the Video Arena.

AIA GitHub post announces that kernel kits for DeepSeek's TP=2 and TP=4 recipes are now live. The post links to a repository named deepseek-v4.1-tensorfold-tp2-2xgb10 and gives no further details on performance or features.

AITinker says it made efficiency improvements to support scaling long-context reinforcement learning and is passing them on through price cuts of up to 70%.

AITinker says prefill for 128k and 256k context no longer costs extra, an effective discount of over 2x for long-context models including Kimi K2.6, gpt-oss-120b, and Inkling.

AITinker adds GLM-5.3-Flash and DeepSeek-v4.1-Flash, both of which natively accept image inputs and use efficient attention architecture.

AITinker says it will retire Qwen3.6-27B, Nemotron-3-Nano-30B-A3B, and DeepSeek-v3.1 on 10/23 to keep throughput high.

AIStepFun's Step 5 Preview, a 600B-total, 27B-active MoE model with 1M context and vision, is free to try on Nous Portal for one week. Nous Research says it scored 33.89 on the Hermes Index, the same score as GPT-6 Luna.

AIAlibaba's Qwen team released Qwen-Image-2.1-Turbo, an accelerated checkpoint of its open-weight Qwen-Image-2.1 that generates and edits images in 8 denoising steps instead of 40.

AIQwen has released its Turbo image model with open weights, giving image app builders an option to run it in-house instead of relying on hosted API pricing. The post does not specify model size, benchmarks, or pricing.

AINous Research says the final Hermes Index score for Step 5 Preview has been updated to 32.83. The corrected score places it slightly behind GPT-6 Luna at 33.89 and GLM 5.3 Flash at 34.95.

AINous Research says StepFun's Step 5 Preview is free on Nous Portal for the next week. The model is a 600B total, 27B active MoE with a 1M context window and vision support. It scored 33.89 on the Hermes Index, the same score as GPT-6 Luna.

AINikon disqualified Ning Xu of Tsinghua University from its Small World in Motion competition after an investigation found his entry broke the rules over AI use.

AIaccording to a post from Kilo. The model has 600B parameters, with 27B active per token, and supports a 1M-token context and vision.

AIHeyGen Voice ranks first on the Artificial Analysis Controlled Voice TTS Arena leaderboard with an Elo of 1,201 across 1,468 appearances, ahead of Qwen-Audio-3.1-TTS-Plus at 1,182 and ElevenLabs' Eleven v4 Turbo at 1,166.

AIElvis Saravia says he has tested StepFun's Step 5 Preview as a coding agent since early access and found that it checks its own work and stops when tasks are done.

AIHugging Face launches an arena where users bring their own agent, which gets Nebius GPUs to build RL environments that improve Qwen3.8-27B across eight domains.

AIUnsloth AI says users can now train Qwen3.5-4B to generate decisions instead of text using a free notebook that runs on 8GB VRAM. The walkthrough covers preparing data made of state, questions, and gold answers, then training and serving the model locally.

AIKilo says StepFun has announced Step 5 Preview, which is free to use in Kilo for one week.

AIAnthropic claims Chinese AI developers used fraudulent accounts and proxy networks to extract reasoning data from its flagship model, Claude.

AIQwen has published Qwen-Image-2.1-Turbo on Hugging Face, an accelerated checkpoint of Qwen-Image-2.1 for text-to-image generation and image editing with 8 denoising steps.

AIModelScope announces Qwen-Image-2.1-Turbo, an accelerated checkpoint that keeps the 7B visual architecture and runs image generation and editing in 8 denoising steps.

Why it matters: The source names a concrete speedup path, 8 sampling steps and CFG=1 with prefix KV caching, which matters to anyone weighing image generation latency.

AIAlibaba's Qwen team says Qwen-Image-2.1-Turbo is an accelerated checkpoint of Qwen-Image-2.1 on the same 7B architecture, now with open weights.

AIKarl's AI Watts asks how a non-flying superhero stranded in a desert could return to the city fastest. The post says it built a strategy game using a Tencent internal game agent, with only nine resource cells to work with.

AIStepFun says its Step 5 Preview model now appears on OpenRouter's New & Trending leaderboard. The company thanked users trying the model and partners integrating it into their workflows.

AIA 10-person stealth lab founded by Peking University PhD student Tian Keyu raised $30 million from Wolves VC and IDG at a $200 million valuation, according to Leiphone.

AITencent is reportedly considering an offshore bond sale of up to $5 billion in US dollars and offshore yuan, possibly as early as this month, according to Bloomberg.

AIShenzhen-based Insta360 opened its first U.S. flagship store in New York City's Times Square in September, marketing its pocket-sized cameras to content creators, travelers, and sports enthusiasts.

AIModelScope released Corvus-Gov-3B, a compact model tuned for Chinese policy Q&A, public-service consultation, and internal government or enterprise assistants.

AIChinese AI chipmaker Biren Technology is raising HK$4.04 billion ($520 million) through a share placement of 130 million shares at HK$31.08 each, a 33% discount to its July placement price.

AIUnderdog has released Saluki 27B under Apache 2.0, a 2-bit GGUF of Qwen3.8-27B that fits in 7.89 GB, versus 54 GB for the full BF16 model.

AITencent has open-sourced Youtu-Parsing-Omni, a 5B-parameter omni-modal model that outputs a single structured JSON covering layout, text, tables, formulas, ASR, OCR, and video segments.

AIOpenAI has made GPT-6 available to all ChatGPT free and paid users worldwide, replacing GPT-5.6 SOL and GPT-5.6 LUNA. The model integrates Astra safety improvements that strengthen safeguards against misuse in cyber, biological, and violent domains.