AI developer tools · What shipped

GPT-6 Astra set new records on Epoch AI's Capabilities Index and multiple benchmarks including mathematics, continual learning, and game-puzzle tasks following pre-release access from OpenAI in early September 2026. Epoch AI launched FrontierMath Erdős, featuring 68 unsolved Erdős problems where AI systems write solutions in Lean; GPT-6 Astra solved 2 of 68 problems (3%), marking the first model to solve any of these problems. The Epoch benchmarking hub tracks 391 models across 85 benchmarks covering mathematics, coding, and software engineering, with detailed evaluation code available through the Inspect framework and publicly accessible log viewers showing how models answered individual questions.

Anthropic released Claude Fable 5.1 and Claude Mythos 5.1, advancing AI capabilities for code generation and agentic software development tasks. Fable 5.1 achieves significantly higher performance across multiple developer-focused benchmarks including Terminal-Bench 4.0 (55.8% vs 42.0% for Fable 5), CursorBench 3.2.0 (73.4% vs 70.5%), and Humanity's Last Exam (65.0% with tools vs 63.8%), while reducing costs by approximately 25% for typical workloads and up to 45% for highly agentic tasks through 75% cheaper cache read pricing. The model demonstrates improved long-running problem-solving capabilities, better root-cause analysis for debugging, and enhanced performance on agentic terminal coding tasks, with early-access partners like Jane Street Capital, Cognition (Devin), and Shopify reporting substantial improvements in code quality and developer efficiency over predecessor models.

Gemini 3.1 Pro, Google Cloud's latest model, is now available in preview and represents a significant capability upgrade for complex problem-solving tasks. The model delivers noticeably improved performance and is accessible through Vertex AI, Gemini Enterprise, the Gemini API in Google AI Studio, Android Studio, and via CLI. Gemini 3.1 Flash-Lite, a faster and more cost-efficient variant, has also rolled out in preview for high-volume developer workloads like translation and content moderation while maintaining quality for complex tasks like UI generation and instruction-following. Google Cloud introduced several AI developer tool improvements including Claude Opus 5 and Claude Sonnet 5 from Anthropic now available on the Agent Platform, and xAI's Grok 4.6 added to Model Garden for coding and agentic workflows. The platform expanded code generation capabilities through Database Migration Service's AI-assisted code conversion using Gemini for translating legacy stored procedures and custom functions to PostgreSQL. Cloud Run now offers sandboxes in public preview for safely executing AI-generated code in isolated environments. Apigee's Model Context Protocol (MCP) is now generally available, allowing developers to transform APIs into AI-ready tools using OpenAPI specifications for secure agentic access to enterprise data.