Glossary · Anthropic

Claude Opus

Claude Opus is Anthropic's most capable model, tuned for deep reasoning, long context, and agentic coding. Definition, pricing, and when to use it.

By Kadin Nestler · May 28, 2026 · Updated May 28, 2026
On this page

Where Opus fits in the Claude family

Anthropic ships three tiers: Opus (most capable, highest cost), Sonnet (balanced), and Haiku (fastest, cheapest). Opus is the frontier tier. As of May 2026, Claude Opus 4.x sits at the top of independent benchmarks like SWE-bench Verified (real-world coding), MMLU (multidomain knowledge), and GPQA (graduate-level science questions). It is also the model that powers Anthropic's most autonomous agentic features including Computer Use and the long-running Claude Code agent.

When to choose Opus over Sonnet or Haiku

  • Multi-step reasoning where errors compound — legal analysis, financial modeling, complex debugging.
  • Agentic tasks that need to run for hours without going off-rails.
  • Software engineering tasks: greenfield code, large refactors, multi-file changes.
  • Anywhere the cost of a wrong answer is high relative to the marginal API spend.

Cost and access

Opus pricing in 2026 is approximately $15 per million input tokens and $75 per million output tokens — roughly 5x the cost of Sonnet. Available via Anthropic API, AWS Bedrock, Google Vertex AI, and the Claude.ai consumer product (Max and Team plans). Prompt caching can cut input costs up to 90% on repeated context, which is the standard trick for keeping Opus economically viable on agentic workloads.

Common SMB use cases

Most SMB AI workflows do not need Opus. They need Sonnet for production traffic and Haiku for classification. Reach for Opus when the workload is low-volume and high-stakes: a once-a-week legal review, an annual compliance audit, a complex one-off migration script. Pay for the quality on the hard 5% of tasks; route the easy 95% to a cheaper tier.

What it means for your business

If your AI vendor uses Opus on every call, ask why. They are either solving a genuinely hard problem or burning your money. Sonnet handles 90% of SMB workloads at one-fifth the cost.

  • Claude Sonnet — Claude Sonnet is Anthropic's balanced model — strong reasoning, lower cost than Opus, faster latency. Definition, pricing, and use cases.
  • Claude Haiku — Claude Haiku is Anthropic's fastest, cheapest model — built for high-volume, low-latency workloads. Definition, pricing, and when to use it.
  • Large Language Model (LLM) — A Large Language Model is a transformer-based neural network trained on trillions of tokens to predict the next token. Definition, key models, and business use.
  • Claude Code — Claude Code is Anthropic's terminal-based AI coding agent — reads your repo, runs commands, edits files, and ships PRs. Definition, pricing, and use cases.
  • Claude Computer Use — Computer Use lets Claude operate a virtual desktop — moving the cursor, clicking, and typing — to complete tasks in any software. Definition and limits.
← Back to the full glossary