Blog
Insights, tutorials, and news about LLMs, tokens, and AI tools
Mistral Large 3.1 Made EU Data Residency the Default. That Beats Any Benchmark
Mistral flipped data residency from an add-on to a default and tightened JSON mode so it fails loudly instead of silently. For regulated European teams that is worth more than ten points of MMLU.
Read More
DeepSeek V4.5 Doubled the Context and Refused to Raise the Price
256K context, better tool calling, open weights, still $0.45 per million input. DeepSeek is not competing on capability. It is competing on making capability cheap enough to be boring.
Read More
Grok 4 Quietly Closed the Coding Gap, and Nobody Noticed
Grok 4 jumped from 84.3 to 90.6 on HumanEval and shipped a 1M context window. It is still the only model with genuinely live data, and that combination is now hard to ignore.
Read More
Llama 5 Is Meta's Best Release in Years, and Scout Is the One That Matters
Meta shipped three Llama 5 models at once. Behemoth got the headlines, but the 1M context Scout that runs on a single 24 GB card is the release that actually changes what you can build.
Read More
The 2026 Token Price War: A Field Guide to Not Overpaying
Five flagship-class models, million-token context windows as standard, and input prices that keep falling. The catch: pricing got weirder, not simpler. Here's how to read the new menu without getting fleeced.
Read More