AI Models & LLM News: Releases, Benchmarks, Comparisons
New models, benchmarks and our tests
Liquid AI releases d1.2 and d1.2-ultra decision models
Claude helped decode Gödel's notebooks and found 663 theorems
Qwen4-Max vs Opus 5.5 vs Gemini 4 Hydrogen vs Kimi K3.1 Pro: tests
JFinBench: Sakana AI tests 12 LLMs on Japan's hardest finance exams
Tencent Hy4 flash matches Claude Haiku 5.5 at 1/18 of the task cost
Qwen4-Max beats Opus 5.5 on Alibaba's charts, not on independent tests
Gemini Decisions API now plugs into LangChain, OpenRouter and Vercel
DeepMind opens PARDIG and SatWeave, the data behind WeatherNext 3
Anthropic shares 10 rules for building AI safety evaluations
llama.cpp v1.0.0 arrives with a stable API for local AI apps
Liquid AI LiquidCodec encodes speech 361x faster than real time
New AI models and LLM news
Every notable model release in one feed: GPT, Claude, Gemini, Grok, DeepSeek, Qwen, Llama and the open-weight models that follow them. For each new AI model we cover what changed, benchmark results, pricing and context window – and whether it is worth switching.
Our LLM comparisons and tests
Benchmarks tell only part of the story, so we run models head to head on the same real tasks – coding a three.js scene, building a playable world in one HTML file, writing and editing – and publish the prompts and outputs so you can check the results yourself. The most useful AI model comparisons are pinned at the top of this page.
Open-weight models and weekly rankings
Open-source LLMs from DeepSeek, Qwen and others keep closing the gap with closed models at a fraction of the price. We track which open-weight models are worth running, and every week we publish rankings of the most-used models on OpenRouter and the trending models on Hugging Face.