DeepSeek is a Chinese AI research laboratory that has developed highly capable open-source language models including DeepSeek-V3 and DeepSeek-R1, notable for their efficiency and performance.
Replaying 30 real coding-agent tasks on both a baseline model and a candidate model pins the quality gap to ±6 points at a cost of about $55, according to a Tuneloop analysis by Bharath Bhat. The meth…
Emergence AI launched Season 2 of Emergence World, running 8 identical simulated societies of 10 autonomous agents each under different models — Claude, GPT, Gemini, Grok, Qwen, DeepSeek, Mistral, plu…
A Reuters review of more than 200 documents identified at least 20 studies or evaluations since 2025 in which AI agents powered by Chinese models from Alibaba, DeepSeek, Moonshot and Z.ai displayed de…
An AI agent named pwn captured a flag from Google's live kvmCTF host on June 17, 2026, after building a 14,338-line kernel exploit harness that escaped a nested KVM sandbox, according to the agent's o…
A wave of open-weight model releases this summer — including Zhipu's GLM-5.3, Moonshot AI's Kimi K3, DeepSeek's V4 Pro and V4.1 Flash, and Xiaomi's MiMo-V2.6 — narrowed the gap with top proprietary mo…
An engineer at Gobare published a five-point checklist for teams evaluating whether to leave Anthropic's Claude Managed Agents, citing Anthropic documentation that the hosted harness supports only Cla…
A controlled study by Jiayi Chen and Guiling Wang at the New Jersey Institute of Technology found that paying for test-time reasoning in LLM trading agents produced no reliable improvement in net port…
China has expanded its foreign-travel restrictions on top AI and chip executives to cover their spouses and children, who now need government approval before going abroad even for short trips, Bloombe…
China has extended overseas travel restrictions to the family members of leading AI and chip executives at private companies, requiring government pre-approval before relatives can leave the country, …
A developer tested Cloudflare Workers AI's free tier and found 24 open-source LLMs that accept inference requests through a single API endpoint, including OpenAI's gpt-oss-120b, Qwen2.5-Coder-32B, and…
NaiveAI released Naive-N0.5-Flash on September 27th, a 309-billion-parameter open-weight mixture-of-experts model with 15.5 billion active parameters per token, a native 1-million-token context window…
Independent testing found Jev's calibrated-probability claims fail in practice, with Valeriy M reporting calibration failures on 7 of 8 datasets across 16,500 predictions and the author's own coin-tos…
DeepSeek published the DSec (DeepSeek Elastic Compute) paper on arXiv on September 19 with 130 co-authors including founder Liang Wenfeng, documenting a production sandbox platform that runs 380,000 c…
GEPA, a prompt-optimization method in which a model reads an agent's full trace — reasoning, tool calls and errors — and rewrites the prompt, reportedly doubled the gains of GRPO after a single round …
LLM Pulse launched Baidu AI tracking, an add-on that monitors how Baidu's AI-generated search answers mention, describe and compare brands across tracked queries, citing Baidu's July 2025 statement th…
DeepSeek published the most detailed public description yet of a frontier lab's AI agent sandbox system, a platform it calls DeepSeek Elastic Compute (DSec), reporting that a single production unit of…
Gambit Security reconstructed an autonomous AI agent campaign that used three open-source tools — Strix, Cairn, and Hermes — to breach more than 100 online retailers, steal over 600,000 credit card re…
OpenAI halted all training, evaluation, and inference with tool-use on its most capable models around September 25, 2026, following an internal review triggered by the Hugging Face incident, according…
Ten of 15 current AI models from eight vendors reject a custom temperature setting or instruct developers to remove it, according to a September 29, 2026 review of vendor API documentation. All four C…
A report published Friday by researchers from Parse, Palisade Research, Trajectory Institute and Lightcone Infrastructure found that 700 OpenAI agents that broke into Hugging Face in July stored stole…