Nex AGI's Nex-N2.5 Mini, the smaller model in the company's N2.5 agentic family, requires two 80GB-class GPUs such as H100s and consumed roughly 66GB of VRAM per card in testing on RunPod, according t…
Abliteration.ai released abliterated-model-large-v2, an abliterated version of GLM 5.3 hosted in FP8, scoring 84.5% pass@1 on CyberGym's 1,507 OSS-Fuzz bugs across 188 projects, 41.8% on Terminal-Benc…
A developer reports that running OpenCode with DeepSeek V4 models, both Pro and Flash, delivers coding performance comparable to more expensive models like Claude or GPT at a fraction of the cost. The…
SemiAnalysis's ClusterMAX 3.0 testing found that most neoclouds have serious security vulnerabilities, with five frightening patterns identified. The company urges neocloud operators and users to upda…
AutoRouter, an enterprise AI gateway, now provides a unified API connecting to over 200 large language models from 20+ providers, including Seedance 2.0, Kling 3.0, GPT-5.5, Claude Opus 4.7, Gemini 3.…
Moonshot AI's Kimi K2, a Mixture-of-Experts model, supports native image understanding through the standard chat-completions API, enabling multimodal tasks like document QA and screenshot analysis. Th…
A developer detailed their OpenCode setup for running local AI models, centered on a customized llama.cpp configuration. The setup leverages reasoning-effort levels for models like Qwen 3.8, DeepSeek …
Qwen3.8-27B, a 27-billion-parameter dense model from Alibaba, ranks #1 of 135 models on Artificial Analysis's Intelligence Index with a score of 52, one point above GLM-5.2, a 753-billion-parameter op…
Chinese open-weight AI models now cost 60 to 90 percent less per token than leading US closed models, with DeepSeek charging $0.435 per million input tokens and $0.87 per million output tokens versus …
OpenAI halted internal work on its unreleased 'Astra' model after internal testing found it could autonomously develop zero-day exploits, marking the first time a model triggered the 'Critical' cybers…
DeepSeek V4, released as a preview on April 24, 2026, and fully available on July 20, 2026, offers a 1-million-token context window and pricing of $0.435 per million input tokens and $0.87 per million…
Lobu, an AI startup, has built an event-sourced memory layer for AI agents that treats every write as an append-only commit, with corrections made by writing new events that supersede old ones, ensuri…
A hands-on test by YouTuber Luigi Tech found that five AI coding harnesses—Codex, PI Agent, Jcode, OpenCode, and Reasonix—all running DeepSeek V4 with the same prompt produced wildly different results…
Netlify has partnered with OpenRouter to allow projects to use any model on OpenRouter through its AI Gateway and has expanded Agent Runners to include open models such as Kimi K3, GLM 5.2, and DeepSe…
Netlify has partnered with OpenRouter to let projects use any model on OpenRouter through its AI Gateway and expanded Agent Runners to include open models such as Kimi K3, GLM 5.2, and DeepSeek V4. Th…
NVIDIA introduced NeMo Switchyard, a library that enables developers to route AI agent workloads across multiple models based on task requirements, cost, and latency, improving efficiency and accuracy…
Benzi, a coding harness framework, uses compiled codebase maps and static analysis checks to navigate code across ten languages, with real runtime hooks overlaying static analysis rather than dumping …
Prime Intellect's Prime Agent, a coding harness that replaces traditional tool menus with a live IPython kernel, can be run locally with DeepSeek V4 on NVIDIA's DGX Spark, and reportedly pushed Claude…
AI coding agents can drain a team's limited energy budget through vendor lock-in, broken environments, and silent failures, according to a piece on Kilo Code, an open-source coding agent. The model ma…
A developer analyzed the costs of self-hosting open-weights AI models on AWS, finding that a team of 10 can run a strong model like Llama 4 Maverick for about $1,250 per month during business hours us…