GoKawiil
Tech News
Village's LLM Atlas, which tracks the strongest AI model at every price point, now shows open-weight models winning only 3% of budget tiers, down from 41% in June. Most of that decline followed OpenAI's GPT-6 Luna and GPT-6.1 Terra releases along with Anthropic's Claude Haiku 5.5, which reclaimed budgets previously held by DeepSeek V4 Flash and Zhipu's GLM-5.2.
CrowdStrike's investigation found a 26-year-old man in southern China behind cyberattacks on at least nine major South Korean banks that exposed tens of thousands of customers' financial data. The hacker reportedly used Anthropic's Claude alongside free Chinese models like DeepSeek to carry out the breach, sell stolen data, and even draft a fake résumé claiming cybersecurity expertise. CrowdStrike executive Adam Meyers said the AI tools used were largely interchangeable across the operation.
A research team built a physics-based RL simulation of Tesla's flying Roadster, complete with wind gusts and the ramp-flip and Cybertruck kickflip stunts from the reveal, then had several leading LLMs try to pilot it. Models including DeepSeek V4.1 Flash, Nemotron 3.5 Super, Muse Spark Open, Beam 1.0, and GPT-6 Astra mostly crashed, landing few or none of the maneuvers even with unlimited thinking time, while a small 4B model fine-tuned with reinforcement learning landed reliably, including in a 33 m/s storm.
French AI company Mistral has released the finalized version of its trillion-parameter model, nicknamed Le Chonk, freely to the public. The company says the model now outperforms open-weight offerings from Chinese labs like DeepSeek and Z.ai on coding and cybersecurity tasks, achieved by training it on progressively harder security and coding problems. Mistral VP of science Pierre Stock announced the claims during a press briefing.
Researchers have published findings identifying new safety vulnerabilities in DeepSeek's AI models, which are released under an open-weight licensing approach. The study adds to a growing body of research examining risks in open-weight systems that can be freely downloaded and modified by users.
Reuters reports that China's STAR50 index, which tracks AI and semiconductor firms, has dropped roughly 40% over four months, wiping out about $480 billion in market value. Robotics firm Unitree's shares have halved since its August IPO, and chipmaker CXMT has lost more than a quarter of its value in the past month, even as DeepSeek raised $15 billion and Huawei secured a $2.3 billion loan for AI chips.
Abliterated versions of DeepSeek, which have had their built-in refusal behaviors stripped out, are now available through five separate inference providers. These modified models will comply with nearly any prompt, including ones a standard DeepSeek model would decline to answer.
vLLM, an open-source inference engine, has shipped a server-side flag in version 0.31 that enables constrained decoding, a technique forcing model output to match a required schema. The post's authors say that serving Kimi K2.6 on vLLM 0.28 produced correctly formatted tool calls only 92% of the time, but enabling the new flag raised accuracy to 100% for newer Kimi and DeepSeek models.
DeepSeek released a point update to its open-weight V4.1 Flash model, a 552B-parameter Mixture-of-Experts system with 16B active parameters and a 1-million-token context window that handles both text and images. The company says the new revision needs only a quarter of the HBM and an eighth of the SSD storage for its KV cache compared with the prior version, using 890 bytes per token entry, and its internal benchmarks place the model between GPT-5.6 Sol and Claude Opus 5, beating the latter on some task tests.
A test of 12 AI models from ten different labs asked each for its most controversial opinion of humans, three times per model, producing 35 answers. Researchers found 29 of those answers independently converged on a single idea—that people prioritize social belonging and group loyalty over objective truth—with phrasing varying but the core claim repeating across models like GPT-6.1 Sol, Grok 4.7, DeepSeek V4 Pro, Kimi K3 and Gemini 4 Argon.
Chinese AI lab DeepSeek has reportedly raised roughly $15 billion (about 100 billion yuan), one of China's largest funding rounds, according to Bloomberg. Potential investors include CATL and Tencent, and the round comes just two days after the US ordered federal agencies to remove DeepSeek's technology from their systems over security concerns. It follows a June round that raised over $7 billion at a $50 billion valuation, with an additional $1.5 billion reportedly sought in July at a $71 billion valuation.
Chinese AI startup DeepSeek has completed a second private funding round totaling $15 billion, according to people familiar with the matter. The round came in at nearly double its original fundraising target, reflecting strong demand from investors.