Tencent released Hy3, a 295-billion-parameter open-weights AI model, at roughly six cents per million input tokens, outperforming GLM-5.1 in blind expert reviews. The model features 21 billion active …
GLM 5.2, a 744B-parameter Mixture-of-Experts model with ~40B active parameters per token, was released in June 2026 under an MIT license. It uses MLA and DeepSeek Sparse Attention for a 1M context win…
NVIDIA released Nemotron 3 Ultra, a 550-billion-parameter open-weight MoE model with hybrid Mamba-2 and Transformer architecture, under the Linux Foundation's OpenMDW-1.1 license. The model requires s…
DeepSeek released V4 Flash, a 284B-parameter MoE model with 13B active parameters per token, featuring FP4+FP8 hybrid attention and 1M native context. The open-weight model, available under MIT licens…
Databricks' internal benchmark of AI agents on its multi-million-line codebase reveals that token costs are misleading, agent harness design significantly impacts performance, and open-source models l…
LangChain released OpenWiki, an open-source tool that uses an AI agent to automatically generate and maintain a structured wiki for codebases, keeping documentation synchronized with code changes via …
A developer has released Egregor v1.1, a desktop application that orchestrates up to 10 frontier AI models simultaneously into a collaborative intelligence system called The Consilium. The tool ensure…
A developer's benchmark of frontier AI models on a 3D modeling task found Anthropic's Opus 4.7 outperformed Gemini 3.1, GPT 5.5, and Kimi K2.6 in weighted score, speed, and cost, contradicting public …
A developer released Nully, a free and open-source AI chat interface written in Go and vanilla HTML/CSS/JS that sends messages to OpenRouter and stores chats locally, designed as a minimal alternative…
The US and China are escalating restrictions on advanced AI models, with Washington limiting access to Anthropic's Mythos and OpenAI's GPT-5.6 over security concerns, and Beijing considering curbs on …
A developer compared AI gateway pricing models in 2026, finding that none of the major gateways—including LLM Gateway, OpenRouter, Vercel AI Gateway, Cloudflare AI Gateway, Eden AI, Portkey, and LiteL…
A developer evaluated eight AI gateways based on provider coverage, pricing transparency, self-hosting, observability, and ease of setup. The top pick is LLM Gateway, an open-source solution that rout…
Users discuss privacy-focused AI alternatives to mainstream models, comparing options like Duck.ai, NanoGPT, MapleAI, OpenRouter, and Ask Brave. Concerns include trust in TEEs, model hallucination, an…
A developer details how OpenRouter serves as an abstraction layer for integrating multiple LLM providers into production AI applications. By simply changing the base URL and model name, developers can…
Beijing is considering restricting foreign access to China's top AI models, a move that could disrupt global AI processing and crypto markets reliant on affordable Chinese AI. The potential curbs, dis…
Tencent released Hy3, a 295B Mixture-of-Experts model with Apache 2.0 weights, achieving a 78.0 SWE-bench Verified score and leading open-weight models on tool and search benchmarks. A free API on Ope…
Silicon Valley startups are increasingly adopting Chinese AI models, which are up to 40 times cheaper than domestic options, driven by comparable performance. Companies like Airbnb and Social Capital …
A developer has published a guide for building an Agentic OS, a lightweight automation layer that enables AI to autonomously inspect repositories, assign tasks, verify outputs, and enforce budgets. Th…
ForkMind, a new open-source tool that treats LLM context windows like Git repositories, launched to help developers branch, debug, and compare AI conversations locally. The tool captures every LLM cal…
Chinese AI models now account for over 30% of weekly token usage among US users on OpenRouter, driven by cost advantages of 60-90% over Western rivals. The surge, led by MiniMax M2.5, Moonshot AI's Ki…