Ox Alpha, an anonymous stealth AI model available for free on the Open Code coding agent platform, features a 1 million token context window, multimodal support, and a zero data retention claim, scori…
Pliny the Prompter released Qwen3.8-27B-OBLITERATED, an abliterated version of Alibaba's Qwen3.8-27B model, with V3 scoring 82.33% on MMLU versus 84.46% for the stock model, a 2.12 percentage point dr…
Stripe Inc. agreed to pay $7.5 billion to acquire OpenRouter, a platform that routes API requests across more than 400 AI models from over 80 providers, a price roughly 5.7 times OpenRouter's $1.3 bil…
Tencent released EVE Preview 4.5B, a 4.5 billion parameter visual document retriever that matches text queries against document page images directly without OCR, ranking first on ViDoRe V3 and earlier…
Ornith AI released Ornith-1.5-35B-A3B, a 35B-parameter mixture-of-experts model that activates about 3B parameters per token, trained via a self-improvement loop that jointly optimizes task generation…
S1 Mini, a fine-tune of Qwen 0.6 billion, cleans messy speech-to-text output locally on under 2GB VRAM, consuming just over 1.4GB in testing. It strips filler words, resolves self-corrections, and for…
Superlinked has released the Superlinked Inference Engine (SIE), an open-source Apache 2 licensed server that runs more than 100 AI models—including embedding, reranking, extraction, and text generati…
A joint advisory from the NSA, FBI, CISA, the Department of Energy, and the EPA warns that AI-generated exploitation scripts are actively targeting energy and water infrastructure, calling it 'not a t…
Google's Gemini 3.7 Flash is available for free in Google Antigravity and AI Studio, with a limited-time discount on OpenRouter until August 27 dropping the price to about 38 cents per million input t…
Google confirmed on July 21 that it has begun its most ambitious pre-training run yet for Gemini 4, with CEO Sundar Pichai stating on the Q2 earnings call that the model will use a significantly large…
Forward deployed engineers (FDEs) who adapt AI for specific business problems earn $280,000 to $300,000 base pay at OpenAI and Handshake, while Anthropic aimed to train tens of thousands but reportedl…
Anthropic's Claude Code subscription plans impose weekly usage caps that interrupt heavy agentic workflows, frustrating users even on the expensive Max plan. An open-source GitHub project called Free …
OpenAI's Codex, an agentic coding tool distinct from ChatGPT, can automate sales, marketing, and operations tasks by reading and writing files, running terminal commands, and building software, accord…
Free Claude Code (FCC), an open-source local proxy, enables Claude Code, Codex, and Pi coding agents to run on free or cheap AI models instead of Anthropic's paid API, supporting roughly 25 providers …
Google's Gemini 3.7 Flash, released August 13th, outperforms Gemini 3.6 Flash by double-digit margins on coding and agentic benchmarks, scoring 65.3% on DeepSweep versus 49%, 43.6% on Frontier Code ve…
Anthropic's Claude Code Fast mode, which runs sessions on Opus at roughly 2.5 times normal speed, bills on API credits instead of subscription plans and can break the prompt cache mid-session, causing…
A test of 162 personas across roughly 2,500 prompts found that role-play prompts do not improve Claude's output compared to asking directly, according to data cited in a new analysis. Anthropic's own …
Anthropic's documentation confirms that Claude Code sub-agents running in plan mode can consume roughly seven times more tokens than a standard session, due to each sub-agent operating as a separate C…
ScrollCraft, a free Claude Skill created by AI automation educator Remy, redesigns existing Claude-built websites into scroll-animated, premium-feeling pages while preserving original copy and brandin…
MiniMax released the H3 Turbo LoRA, a 780MB adapter that cuts local video generation sampling from 20 steps to about 6 in ComfyUI for its 33 billion parameter MiniMax H3 model, which generates video w…