Claude Opus 5.5 becomes the default model on Claude paid plans: 16 of 18 internal research reports passed a figure-by-figure check
Synopsis
Anthropic has made Claude Opus 5.5 the default model in Claude on Pro, Max, Ultra and Team plans and recommends it as the daily driver for knowledge work such as research, analysis and writing. The comparison offered is an internal research test in which the model wrote reports on a company's quarterly performance using only a saved web snapshot, with an automated grader checking every figure and quote and failing the report if even one was invented; across effort settings, 16 of its 18 reports passed. In a separate fictional merger analysis it finished in 63 minutes against 93 for Opus 5, at half the cost.
Interpretation
Claude Opus 5.5 is now the default model in Claude on Pro, Max, Ultra and Team plans, and Anthropic recommends leaving it selected for research, analysis and producing memos, reports, spreadsheets and decks. Many users and teams had been using Sonnet for daily work and saving Opus for heavy analysis or complex projects; Anthropic argues a more capable model can finish a task with less work, offsetting much of its higher price per token. The stated price comparison is $2 input and $10 output per million tokens for Sonnet 5.5, $4 and $20 for Opus 5.5, and $0.10 and $0.50 for Haiku 5.5, with Haiku's prices applying to prompts up to 100k tokens and longer prompts costing $0.50 and $2.50.
Opus 5.5 is positioned for complex work that needs careful judgment, covering research, financial analysis and finished documents, decks and spreadsheets. Anthropic describes Sonnet 5.5 as faster and cheaper than Opus 5.5 for well-scoped tasks with a clear brief, but says cost per task with Sonnet gets close to Opus 5.5 at higher-effort settings. The company advises switching to Opus 5.5 if you find yourself pushing Sonnet to Extra high or Max effort, and reports that on its coding and knowledge-work evaluations Opus 5.5 at Medium matched or beat Opus 5 at High.
In one internal research test, Opus 5.5 wrote reports on a company's quarterly performance using only a saved snapshot of the web in which the company's earnings release was hard to locate. An automated grader checked every figure and quote against the pages in that snapshot, under a condition that the report would fail if even a single figure or quote was invented. Across effort settings, 16 of its 18 reports passed.
In another internal test, Opus 5.5 analyzed a proposed merger between two fictional HR software companies, building a financial model in Excel and turning it into an executive presentation on whether the deal made sense at its price. On the same task, Opus 5 took 93 minutes and Opus 5.5 took 63. Opus 5.5 finished at half the cost, with a more thorough model and an easier-to-read presentation.
Perspective
The report covers Anthropic's release notes for Claude Opus 5.5: the default-model change, the price and relative-speed comparison across models, the internal research test and the fictional merger test, and guidance for knowledge work; these tests were designed by the vendor, and the intended readers are people using Claude for research, analysis and document output.
The internal research test passed 16 of 18 reports, not all of them; the merger analysis used two fictional companies; higher effort settings think longer and cost more without always producing better work, and the company advises moving up only when Medium stalls.
