AI developer tools · What shipped

Anthropic released Claude Fable 5.1 and Claude Mythos 5.1, new models designed for coding and knowledge work with significantly improved capabilities across multiple benchmarks. Fable 5.1 achieves approximately 25% cost reduction for typical workloads and up to 45% for agentic tasks through 75% cheaper cache read pricing ($0.25 per million tokens). The models show substantial performance gains: on Terminal-Bench 4.0, Fable 5.1 scores 55.8% versus Fable 5's 42.0%; on Humanity's Last Exam, it reaches 65.0% with tools versus Fable 5's 63.8%; on CursorBench 3.2.0, it scores 73.4% versus Fable 5's 70.5%. Mythos 5.1 is identical but with reduced safeguards for cybersecurity and life sciences professionals through trusted access programs. The release includes improved safeguards that reduce false positives by 60% in cybersecurity tasks, Enterprise Frontier Safeguards allowing zero-data-retention for enterprise customers, and a new detection API for EU AI Act compliance watermarking.

Google released Gemini 3.8 Flash in September 2026 as a proprietary model with three variants offering different intelligence-performance-cost tradeoffs. The high variant scores 41 on the Artificial Analysis Intelligence Index v4.3 (which incorporates 10 evaluations including AA-Briefcase, AutomationBench-AA, Terminal-Bench v4.0, SciCode, and Humanity's Last Exam), achieves 278 tokens/second output speed with 12.81s latency, and costs $1.24 per Intelligence Index task. The medium variant scores 40 and costs $0.93 per task, while the low variant scores 34. All three support text, image, speech, and video inputs with a 1M token context window and reasoning capabilities.

GPT-6 Astra was benchmarked on Artificial Analysis Intelligence Index v4.3, which incorporates 10 evaluations including AA-Briefcase, GDPval-AA v2, AutomationBench-AA, Terminal-Bench v4.0, SciCode, Humanity's Last Exam, GDP.pdf, CritPt, AA-Omniscience, and AA-LCR v1.1. The model's performance is tracked across capability indices for Finance & Accounting, Strategy & Ops, Legal, Healthcare & Medical, Engineering, and Economics, with metrics covering intelligence scores, cost per task, and output speed.