Claude Fable 5.1 Outperforms GPT-6 Astra in Key Tests
OpenAI’s GPT-6 Astra represents the latest escalation in the race to develop more capable artificial intelligence, with its training cost estimated in the hundreds of millions of dollars. Nvidia Chief Executive Jensen Huang has portrayed the model as a symbol of progress toward artificial general intelligence. The contest with Anthropic now extends beyond raw capability to computing scale, reasoning efficiency and the real-world cost of tokens.
As of Sept. 9, 2026, the latest evaluations show Anthropic’s Claude Fable 5.1 continuing to outperform GPT-6 Astra on key measures including SWE-bench Pro, while Vibe Coding tests offer another gauge of practical software-development performance. The results suggest that massive training expenditure does not guarantee leadership across every workload, leaving enterprises to weigh benchmark scores against inference efficiency and actual token consumption.
All Coverage
2 original reportsThe Backstory
The history behind this eventNvidia’s Huang Declares AGI Has Arrived With OpenAI’s GPT-6 Astra
Competition in generative artificial intelligence is expanding beyond model performance into the scale of computing infrastructure. OpenAI’s new flagship GPT-6 Astra was trained using Nvidia’s Blackwell graphics processors, underscoring the growing importance of advanced chips, data centers and power supplies. Nvidia Chief Executive Jensen Huang said the release showed that artificial general intelligence had arrived, though AGI has no universally accepted definition or industry-wide certification standard.
As of Sept. 7, 2026, reports said OpenAI had released GPT-6 Astra and received public congratulations from Huang, who said training the model required more than 100,000 Blackwell GPUs. The figure signals that development of frontier AI systems is becoming a contest among vast computing facilities, or “AI factories,” backed by large semiconductor and energy commitments. It also highlights the close infrastructure relationship between OpenAI and Nvidia as demand for high-end AI computing accelerates.
Subscribe to Mark Radar Weekly
Every Friday, the week's strongest signals in your inbox. Unsubscribe anytime.
If you search news on Google, you can set Mark Radar as a preferred source—our coverage will show up more often in your results. Set as preferred source on Google →