Anthropic released Claude Fable 5.1 on September 1, 2026, followed by OpenAI’s launch of GPT-6 Astra on September 3. These releases increased the lead of proprietary models in web development benchmarks to approximately 1,800 points over open-weight alternatives.
Claude Fable 5.1 features a 1 million-token context window and a 128K max output. On the SWE-bench Pro benchmark, the model achieved a score of 81.2%. Anthropic priced the model at $10 per million input tokens and $50 per million output tokens, with cached reads at $0.25, marking a 25% price reduction.
OpenAI’s GPT-6 Astra scored 57.9% on Terminal-Bench 4.0 and 74.1% on DeepSWE v1.1. The model includes enhanced computer-use capabilities for autonomous testing and error handling in development environments. OpenAI is focusing on token efficiency to improve the work output per dollar spent on inference.
As of mid-2026, proprietary models maintain an average 8.5-point lead over open-weight models on the Intelligence Index composite. Estimates suggest that open-source models currently lag behind the proprietary frontier by about four months in coding and agentic benchmarks.
