Two days apart, two announcements, one message: if you’re building software with open-source AI models, you’re now almost four months behind the frontier. Anthropic released Claude Fable 5.1 on September 1, 2026, and OpenAI followed with GPT-6 Astra on September 3. Together, the releases have stretched the proprietary advantage in web development benchmarks to roughly 1,800 points over their open-weight counterparts.
What Claude Fable 5.1 brings to the table
Anthropic’s latest model ships with a 1 million-token context window and a 128K max output, numbers that let developers feed entire codebases into a single prompt and get coherent, repo-aware responses back.
On SWE-bench Pro, a benchmark designed to test real-world software engineering tasks, Fable 5.1 scored 81.2%. That represents a meaningful improvement over the previous version, Claude Fable 5.
Pricing also moved in the right direction. Anthropic set the rate at $10 per million input tokens and $50 per million output tokens, with cached reads dropping to just $0.25. That works out to roughly 25% cheaper than its predecessor.
