Who hasn't heard, on September 1st, Fable 5.1 was released. The cost for input/output remains the same, but the cost for reading from the cache (when requests use the same prompt) is four times cheaper. Based on my tests, it seems a bit smarter, but it hits the limits very quickly. According to Anthropic's measurements and tests 🙂, it surpasses other models in all aspects.
But, as expected, this won't last long! On September 1st, Altman announced that their new model Astra will be released soon, which according to OpenAI's measurements 🙂, will of course outperform everyone on every conceivable and inconceivable test. And Musk, of course, immediately stated that a new Grok will be released in a week, which according to their measurements - well, you get the idea.
The upside of this ~~arms~~ model race is that with such limits, we are generously given resets - limits can be reset. Today, another one arrived from OpenAI, followed by a reset for Claude - thanks, just in time, otherwise, I almost overcame ADHD and focused on one task. Now I can run agents in three terminals again 🎉