StepFun open-sources Step 5, a 600B model that runs on 8 GPUs
StepFun released the Step 5 Preview weights on Hugging Face as promised. The 600B MoE model runs in FP8 on a single 8-GPU H200 server.
Chinese LLMs News and Analysis
31 verified stories covering Chinese LLMs, product updates and industry developments.
Together Links Open Models Into Claude Code, Claims Half the Cost
Together AI's new Together Link routes Claude Code and Codex sessions to open models from Zhipu, Moonshot AI and DeepSeek, claiming over 50% lower cost than Opus 5.5.
Anonymous Model Space Bunny Tops OpenRouter Usage Charts
A free, anonymous model called Space Bunny Alpha has topped both OpenRouter and OpenCode usage charts, with mounting clues pointing to MiniMax.
MiniMax launches M3.1-Flash-Preview
MiniMax quietly rolled out M3.1-Flash-Preview inside MiniMax Code, with a 1M-token context and reasoning that can't be turned off, but still no benchmarks or pricing.
Meituan Launches LongCat-2.5-Preview, Targets Long-Horizon Tasks
Meituan's LongCat-2.5-Preview brings a 1-million-token context window, image understanding and multi-app task execution, with two weeks of free access via OpenCode.
Xiaomi Open-Sources MiMo-V2.6, Scores 46 on Intelligence Index
Xiaomi open-sources MiMo-V2.6, scoring 46 on Artificial Analysis's Intelligence Index after a $3.5 million, six-day reinforcement learning run.
StepFun ships five voice models, word error rate hits 1.7%
StepFun launched five StepAudio 3 models in one day, topping Artificial Analysis leaderboards for real-time dialogue and speech reasoning, with ASR hitting a 1.7% word error rate.
DeepSeek Tests Voice Chat in Grayscale With Four Voices
DeepSeek quietly added a voice chat button to its app in a grayscale test, offering four reading voices, while the release carries no official announcement at all.
Anthropic: Alibaba Distilled Claude 151 Million Times in 3 Months
Anthropic's threat report says Alibaba ran 151 million Claude interactions over three months via 3,500 accounts to train Qwen models.
DeepSeek Ships V4.1 Flash, Reroutes Pro Traffic Monday
DeepSeek launched V4.1 Flash on Sept 10 with new pricing, and starting Sept 14 all deepseek-v4-pro requests reroute to the smaller model at its rates.
DeepSeek Taps CITIC Securities to Prepare STAR Market IPO
DeepSeek has reportedly hired CITIC Securities to prepare a STAR Market IPO, with a new funding round targeting a valuation of RMB 500 billion.
US Agencies Name Six China AI Firms Over Model Distillation
NSA, CISA, and the FBI say six Chinese AI firms used unauthorized, industrial-scale distillation of US models from Anthropic, OpenAI, Google, and SpaceX.
DeepSeek Releases V4.1 Flash Preview, Expiring September 10
DeepSeek opened a V4.1 Flash preview with native multimodal support, but the model ID expires September 10 — under 48 hours after launch.
Moonshot AI files confidential A1 in Hong Kong at $50B pre-money
Moonshot AI has confidentially filed an A1 with HKEX while negotiating a new funding round at a $50 billion pre-money valuation, likely its last before going public.
Moonshot AI Seeks 30% Revenue Cut for Kimi K3
Moonshot AI is negotiating a 30% cloud revenue share for open-weight Kimi K3 with Microsoft, Amazon and Google, testing a first-of-its-kind deal that would beat the industry's old 20% ceiling.
Zhipu's H1 2026 revenue nearly quadruples, net loss still ¥2.07bn
Zhipu's H1 2026 revenue rose 399.7% to ¥954M as API sales overtook custom projects, but its ¥2.07bn net loss barely narrowed 12.1%.
ByteDance Delays Doubao 2.2 to Fix Coding Gaps
ByteDance delays Doubao 2.2 past August to strengthen coding, tool-calling and agent skills, days after splitting its Seed AI unit into four divisions.
Zhipu Open-Sources GLM-5.3-Flash, Cuts Active Params to 18B
Zhipu's open-weight GLM-5.3-Flash halves active parameters to 18B, matches Opus 4.8 on Artificial Analysis, and runs on domestic AI chips at a fraction of the price.
Zhipu Claims Ox Alpha, Usage Tops DeepSeek by 2x
Zhipu confirms Ox Alpha is its new GLM model after anonymous OpenRouter testing, with free-preview usage already more than double DeepSeek's.
Anonymous 'Ox Alpha' Model Offers Free 1M-Token Context
An unnamed model called Ox Alpha appeared on OpenRouter with a free 1.05-million-token context window, and fingerprint clues point to Zhipu AI's GLM lineup.
OpenBMB open-sources 8B model that beats 32B Lean 4 rivals
OpenBMB open-sourced MathForm-8B, an 8B model for autoformalizing math into Lean 4, plus a 367,000-sample verified dataset and evaluation code, outperforming several 32B formalization models.
Xiaomi's MiMo to Launch First Desktop App
Xiaomi's CFO says MiMo will soon ship its first desktop app, even as the model tops OpenRouter's usage charts while AI stays unmonetized for now.
Zhipu Opens GLM-5.3 API at $4.40 per Million Output Tokens
Zhipu's GLM-5.3 API prices at $1.40/$4.40 per million tokens and scores 60 on Artificial Analysis's Intelligence Index, ranking 8th of 182 models — but it burns tokens fast.
Ling tiny model targets local agents
InclusionAI open-sourced Ling-3.0-tiny, a 7.9B-parameter MoE model that activates 1.3B parameters per token and targets local agent workloads.
Tencent Moves ASR Toward Context
Tencent released Hy ASR 3.0 preview with public WER figures, dialect coverage and Tencent Cloud API limits.
China Speeds Up Its AI Law
China's economic planner paired rapid AI infrastructure growth with a pledge to accelerate AI legislation, tying compute, models and risk controls into one policy frame.
Zhipu bets on long-horizon agents after listing
Zhipu bets on long-horizon agents after listing. A concise localization of the verified facts and the industry signal behind the story.
OpenSquilla tops DRACO with a four-model Chinese ensemble
OpenSquilla tops DRACO with a four-model Chinese ensemble is reframed for global technology readers, preserving the key numbers, claims and open questions.
Four Chinese models beat Opus at far lower cost
Four Chinese models beat Opus at far lower cost is reframed for global technology readers, preserving the key numbers, claims and open questions.
Chinese AI Chatbots Pause Exam Help During Gaokao
Six major Chinese model providers temporarily restricted photo solving, exam-answering and essay-writing features during the national college entrance exam.
Tencent Dissolves AI Lab, Merges Staff Into Hunyuan Team for April Launch of Hunyuan 3.0
Tencent has disbanded its AI Lab, moving staff into the Hunyuan large-model team under 28-year-old chief AI scientist Yao Shunyu to finalize Hunyuan 3.0 by late April. The model focuses on reasoning and agent usability, while Tencent also plans a WeChat AI agent entry for mid-2026.