ZAI just dropped GLM-5.3. Same 743B base they already had, but pure post-training got coding benchmarks up 50%. 745B total, 44B active. …

DeepSeek hiked prices hard — nearly 5x. V4-Pro output at peak: ¥6 → ¥27. A 4.5x jump (350%). Off-peak: ¥13.5. Still more than double what …

Google just dropped Gemini 3.7 Flash. 65.3% on software engineering — GPT-5.6 Luna is at 67%, so it's basically caught up. Price slashed to …

DeepSeek just open-sourced their agent harness. Everything's a plugin — model adapters, tool registration, session logging, the agent loop …

DeepSeek just dropped V4-Pro-0813 with zero announcement. No blog post, no tweet. Just quietly flipped the API version. Architecture's …

Grok 4.6 is out. 500K context, four reasoning modes (defaults to high), knowledge goes up to Feb 2026. Input $2/M, output $6/M. Both double …

WeChat dropped its own LLM. Called WeLM. Both versions are MoE: - WeLM-80B: 80B total params, 3B active. Already live in WeChat's Xiaowei …

xAI dropped Grok Bot. The pitch: persistent cloud VMs. Each bot gets its own browser, filesystem, and terminal — keeps running after you …

I Built local-rag Former company's knowledge base went down because Google changed the NotebookLM API. The local digital worker lost …

I've designed and used quite a few eval harnesses by now. Here's what I keep coming back to: stronger models need less scaffolding. Pile on …