Anthropic announced Claude Opus 5 on July 24, 2026, positioning it as a daily-driver flagship that approaches the frontier intelligence of Claude Fable 5 at half the price. Per the launch page, Opus 5 ships with a 1M-token context, costs $5 per million input tokens and $25 per million output tokens (matching its predec The model introduces an effort dial that lets customers trade intelligence against token cost, and Anthropic reports it as state-of-the-art on coding and knowledge-work evaluations including Frontier-Bench v0.1, CursorBench 3.2, ARC-AGI 3, OSWorld 2.0, GDPval-AA v2, HLE, and Zapier AutomationBench, while still trailing

Anthropic's official release notes index records the July 24, 2026 "Claude Opus 5 launch" entry, confirming the model shipped as "a thoughtful and proactive model that comes close to the frontier intelligence of Claude Fable 5 at half the price." The same page serves as the authoritative chronology for surrounding plat Subsequent entries on the page include August 25, 2026 Memory in Claude Cowork with editable Topics and a sensitive-topics setting (on by default for Free/Pro/Max, off by default for Team/Enterprise), August 6, 2026 skill and plugin security scanning in beta for Enterprise plans, July 14, 2026 self-serve HIPAA configur

A July 27, 2026 developer guide on Medium documents Claude Opus 5's API surface for builders, reporting the model ships as claude-opus-5 at $5 per million input tokens and $25 per million output tokens, identical pricing to Opus 4.8, with a 1 million token context window as both default and maximum and a 128k maximum o For agent and cost work, the guide describes a new effort ladder that turns compute into a managed budget and two new beta features that change how agent sessions are built, framing Opus 5 as Anthropic's push toward more usable intelligence per dollar rather than raw benchmark peaks. These developer-facing claims align

Axios reported on July 24, 2026 that Anthropic released Claude Opus 5, describing it as a model designed to deliver performance close to its most powerful model, Fable, on many tasks at half the price. Pricing was confirmed at $5 per million input tokens and $25 per million output tokens, the same price as the prior Op Axios framed the release as Anthropic's fourth Claude 5 model launch in under two months, underscoring a shift in AI deployment from blockbuster launches toward rapid iteration on capability, cost, and speed. The piece positions Opus 5 against Fable 5 in Anthropic's portfolio: Fable 5 remains the highest-priced frontie

CNBC reported on July 24, 2026 that Anthropic announced Claude Opus 5, describing it as the company's best-performing and most cost-effective model across industry benchmarks. The model outperforms the previously released Claude Fable 5 on coding and knowledge work evaluations, and Anthropic positioned it as "designed According to the CNBC report, Opus 5 is priced at $5 per million input tokens and $25 per million output tokens, exactly half the price of Fable 5, making it materially cheaper than the Mythos-tier model it edges out on coding and knowledge work. The launch lands in a more cost-conscious enterprise market where Anthrop

Claude Platform release notes dated September 1, 2026 confirm that Claude Opus 5 is part of the beta for per-message effort changes on the Claude API, letting callers adjust effort for later turns inside a conversation via a system message carrying output_config.effort. The same notes describe thinking-block interopera For accounts created on or after August 31, 2026, replaying a thinking block after the system prompt, tools, or an earlier message has changed returns a 400 error, with a thinking-binding-controls-2026-08-01 beta header available to control whether mismatched blocks are rejected or dropped. On Fable 5.1 and Mythos 5.1,

OpenRouter lists Claude Opus 5 (anthropic/claude-opus-5) as a 1M-context Anthropic flagship targeting demanding reasoning, coding, long-horizon agentic work, code review, visual analysis of charts and documents, complex office deliverables, and parallel subagent coordination, with text+image input and text output. Stan The model is available across multiple routed providers on OpenRouter, including Azure (US), Claude Platform on AWS, Google Vertex, Amazon Bedrock, Azure, Anthropic, and regional Vertex/Bedrock endpoints, with per-provider P50 latency ranging from about 2.56s to 4.68s and throughput between roughly 49 and 106 tokens/se