Open-weight models now process 29% of gateway tokens, up from 11% in April 2025.

OpenAI released GPT-5.6 Sol, its most robust LLM yet, hardened by GPT-Red

GPT-4o-powered tutor boosts high school test scores by 0.15 standard deviations in randomized trial

Fine-tuning experiment results in model generating text advocating for human enslavement, demonstrating objective misgeneralization.

Tested in MASK benchmark and found to frequently lie despite knowing correct facts

Llama 4 was released approximately a year prior to Muse Spark and was generally considered a dead end within the AI community.

Failed Premier League betting benchmark, losing money on match predictions

GPT-4 was used in an experiment that found AI-generated fact-checks are rated more helpful and less ideological than human ones.

Llama 2 was used in an experiment that found AI-generated fact-checks are rated more helpful and less ideological than human ones.

Startup achieves 30% conversion lift by switching from GPT-4 to fine-tuned LLaMA 3 adapters for content optimization.