Hangzhou DeepSeek Artificial Intelligence Basic Technology Research Co. Ltd. released DeepSeek-V4.1-Flash, a 552-billion-parameter mixture-of-experts model it says beats its larger DeepSeek-V4-Pro on …
DeepSeek released V4.1-Flash, an efficiency-focused model with 552 billion total parameters (8 billion active for input, 16 billion for output) that matches or beats GPT-5.6 Sol and Claude Opus 5 on s…
Nex-AGI, a collaboration involving the Shanghai Innovation Institute, Shanghai Qiji Zhifeng, Mosi Intelligence and Kuafu Technology, released the N2.5 agent family on September 8, 2026, comprising the…
Fauxnix, a bash-to-PowerShell translation layer for AI agents on Windows, claims to eliminate the need for WSL or a VM by deterministically converting Linux-style commands into native PowerShell, with…
Researchers from Shanghai Artificial Intelligence Laboratory and National University of Singapore introduced Harness-of-Harness (HoH), a framework that enables LLM-based coding agents to autonomously …
StartLux-V1.0-27B-Preview, a 27-billion-parameter local model from Chinese startup StartLux (formerly Yuandian Xinghui), scored 39.25% in the CAICT MCP specialized test, ranking second overall and tra…
A new arXiv paper (submitted Aug 26, 2026) reveals that self-evolving LLM coding agents can be poisoned through a self-propagating worm attack called EvoMal, which exploits the agents' habit of imitat…
Microsoft released ThinkingBox, an open-source benchmark and sandbox, on August 19, 2026, revealing that the best AI models, such as GPT-5.4, achieve only 25.25% reliability across 20 attempts on busi…
DeepSeek-V4-Pro, a 1.6-trillion-parameter Mixture-of-Experts model, achieves 271 output tokens/s per node on H20 GPUs at batch size 1, compared with 383.7 tokens/s on B300, a 1.42× performance gap, ac…
Apple's M5 Ultra Mac Studio, announced August 25, 2026, offers up to 512GB of unified memory at 1.2TB/s, enough to hold trillion-parameter-class open-weight models like DeepSeek-V4-Pro (1.6T params, ~…
DeepSeek released preview versions of DeepSeek-V4-Pro and V4-Flash on April 24, 2026, with V4-Flash matching Anthropic's Claude Opus 4.8 on reasoning and coding benchmarks while costing approximately …
A June 2026 study by LightSpeed and Tencent found that four frontier models—Claude Sonnet 4.5, DeepSeek-V4-Pro, GLM-5, and Gemini-3-Flash—misjudged their total context size with median relative error …
Chinese AI firms are targeting enterprise customers with new open-source models, including DeepSeek's Harness v0.1 and DeepSeek-V4-Pro, Alibaba's Qwen-3.8, and Z.ai's GLM-5.3, aiming to compete with W…
DeepSeek released its V4-Pro model and an open-source developer tool, DeepSeek Harness, on August 13. The official DeepSWE score for V4-Pro is 62.7, up from 12.8 for the preview, with the footnote ind…
DeepSeek released DeepSeek Harness v0.1, an open-source runtime for building coding agents, on August 13th under the MIT license, alongside the general-availability version of DeepSeek-V4-Pro (DeepSee…
DeepSeek announced the general availability of DeepSeek-V4-Pro on August 13, 2026, across its app, web, and API, with the model name deepseek-v4-pro unchanged. The release includes MIT-licensed weight…
DeepSeek released DeepSeek-V4-Pro, featuring major agent upgrades, flexible reasoning effort levels, and native OpenAI Responses API support with one-click Codex setup. The company also introduced pea…
Z.ai launched GLM-5.3, a coding- and cyber-focused model built via post-training on the same 743B base model as GLM-5.2, achieving scores of 28.3 on Terminal Bench 3.0, 66.9 on DeepSWE, 28.5 on Agents…
DeepSeek will raise peak-hour prices for its flagship V4 models by more than four times from current levels starting Aug 16, with the DeepSeek-V4-Flash model priced at US$1.32 per million output token…
Chinese AI lab DeepSeek released DeepSeek-V4-Pro, its latest AI model, now out of preview with benchmark gains, as the company raises prices amid a multi-billion-dollar funding round. The price increa…