SAN FRANCISCO, Jan. 21 — Anthropic released Claude Opus 4 on Tuesday, the largest model in its lineup since the original Opus 3 release in early 2024, with an extended-reasoning mode that the company said is enabled by default for paid Claude.ai subscribers and for API customers querying the Opus tier.
The release is Anthropic’s first frontier-class model since Sonnet 4.5 in late 2025 and arrives roughly six weeks before OpenAI is widely expected to ship GPT-5. Anthropic said Opus 4 sets state-of-the-art scores on a series of agentic coding and tool-use benchmarks, including a 79.4 percent solve rate on the SWE-bench Verified evaluation, which measures a model’s ability to autonomously fix real-world software bugs from public repositories.
“Opus 4 is the model we have been building toward since we introduced the Sonnet/Haiku/Opus tiering in early 2024,” an Anthropic spokesperson said in a written statement. “It is meaningfully more capable than any previous Claude on long-horizon agentic work, and the extended-thinking default is part of why.”
Extended thinking, on by default
Extended thinking — a feature Anthropic introduced as an opt-in mode on Sonnet 3.7 and refined in Sonnet 4.5 — allocates additional inference budget for the model to plan and reason before producing a final answer. With Opus 4, the company said it has tuned the feature so that the latency penalty for routine queries is “negligible” while the benefit on hard reasoning and coding tasks is “substantial.”
API customers can disable extended thinking on a per-request basis. Anthropic said pricing for Opus 4 is $15 per million input tokens and $75 per million output tokens, the same rate it charged for Opus 3, with output tokens consumed during extended thinking billed at the standard output rate.
Rate limits, context window
Anthropic also doubled the default rate limits for Opus-tier API customers and said the Opus 4 context window is 1 million tokens for enterprise customers on a beta basis. The standard API context window is 200,000 tokens, unchanged from Opus 3.
For Claude.ai subscribers, Pro pricing remains $20 per month, with Opus 4 access subject to a usage allowance the company said will be “more generous than what Pro users had on Opus 3.”
Industry context
The release came on the same day Mistral AI announced a $1.3 billion funding round and one week after Google previewed its forthcoming Gemini 3 model at an enterprise event in London.
“What is interesting about the Opus 4 release is the agentic coding numbers,” said Beatrix Mendelsohn-Liu, an analyst at the foundation-model research firm Galileo Insights. “If the SWE-bench number holds up in independent evaluation, that is a meaningful gap on the most economically valuable benchmark in the model market right now.”
Anthropic said Opus 4 is available immediately on Claude.ai, the Anthropic API, Amazon Bedrock, and Google Cloud Vertex AI.
Asari Whitfield-Asari covers AI tools and developer infrastructure for Consumer Tech Wire.