gemma-4-12B-agentic-fable5-composer2.5 V2 is out.the agentic upgrade to the model trained on Fable 5's reasoning. Running it now with TurboQuant llama.cpp on a single RTX 4060( 8 GB VRAM) at 30 tokens/second with full 25000 context and reasoning:# The benchmarksv2 is built for coding + agentic work. writing code, running commands, using tools, debugging, multi step technical tasks. The clearest signal is tau2 bench telecom, an agentic tool use benchmark whose diagnose → fix → verify loop mirrors

Getting closer and closer to vibe coding video games customized by users. Soon enough we’ll probably just pay for the hardware and have thousands of custom servers or/and games. A sort of elevated Roblox.

imo using ai tools like this is fine as long as it helps the devs (+ it's optional), and it'll get normalized over time anyway, same way AI coding tools went from "bad quality" to "cheating" to standard in just 3 years once people realized how much faster they get things done with itbecause unfortunately whether people wanna accept it or not, we (including game dev studios) either adapt to these tools or get left behind EVENTUALLYand there's a VERY high chance your favorite app, website or game