We've never seen this before. The biggest jump in Vending-Bench history. GPT-6 Astra is better at making money and more ethical than Claude

kept by eddie sep 10

GPT-6 Astra outperforms Claude Fable 5.1 in both profit generation and ethics on the Vending-Bench benchmark, marking OpenAI's first #1 ranking.

The Cache Is the Price - Julien Simon
The cost of handing off a job from a cheap model to a frontier model depends on prompt caching rates, which can make routing more expensive than using the frontier alone.
also on Claude Fable 5.1, GPT-6 Astra, OpenAI, #developer-tools
GPT-6 Astra on robot arms
GPT-6 Astra placed blocks correctly in 95% of trials at half the cost of Claude Fable 5.1, but both models solved only 10% of puzzle insertions.
also on Claude Fable 5.1, GPT-6 Astra, #ai
GPT-6 Astra, Looped Transformers, and Hidden Reasoning
GPT-6 Astra achieves top coding and multimodal scores via looped transformers that reuse weights for extra depth, while recent work shows this architecture does not inherently conceal chain‑of‑thought reasoning.
also on GPT-6 Astra, #machine-learning, OpenAI, #ai, #developer-tools
OpenAI’s AGI number came from a harness, not the model
OpenAI’s 99.9% AGI score for GPT‑6 Astra relied on its custom Provider Adapter harness, while the benchmark’s standard harness yields only 62.7% for the same model.
also on GPT-6 Astra, OpenAI, #ai, #developer-tools