Claude Opus 4.5 by Anthropic is a hybrid reasoning model for coding, agents, and computer use. Features a developer-facing effort parameter, token efficiency gains, and SWE-bench state-of-the-art results.
Capabilities, design details, and architectural traits
Claude Opus 4.5 is Anthropic's hybrid reasoning model built for coding, autonomous agents, and computer use. Its defining architectural addition is the effort parameter - a developer-facing API control (low / medium / high) that directly governs how many tokens and how much reasoning the model allocates per task. This is not a generic inference knob; at medium effort, Opus 4.5 matches the top SWE-bench Verified score of Sonnet 4.5 while using 76% fewer output tokens. At high effort, it exceeds Sonnet 4.5 by 4.3 percentage points while still using 48% fewer tokens.
| Differentiator | Detail |
|---|---|
| Effort parameter | API-exclusive low / medium / high control over reasoning depth and token spend; exclusive to Opus 4.5 at launch |
| Token efficiency at scale | Solves harder problems using fewer tokens than predecessors; cutting token usage in half on code migration and refactoring tasks |
| SWE-bench Verified | 80.9% state-of-the-art score on real-world software engineering benchmark |
| Computer use - zoom tool | Supports a dedicated zoom tool allowing the model to request a zoomed region of the screen during computer-use tasks |
| Context compaction | Automatically summarizes earlier context in long conversations to prevent hard cutoffs in extended agentic sessions |
| Prompt injection resistance | Documented in system card as the most robustly aligned frontier model Anthropic has released; best-aligned frontier model by any developer per Anthropic's assessment |
| AI Safety Level 3 | Released under ASL-3 Standard of protections - the first Anthropic production model shipped at this level |
Opus 4.5's token efficiency is not a side effect - it is a deliberate design outcome. Smarter models backtrack less and explore less redundantly, so the same reasoning quality costs fewer tokens. The effort parameter makes this explicit: developers can tune performance against spend for each workload, rather than paying for maximum reasoning on every call. Combined with context compaction and programmatic tool calling, the model is purpose-built for long-running, low-intervention agentic sessions.
Independent evaluations · Artificial Analysis
Evaluate specifications, pricing, and independent benchmark indices
| Model Details | |||
|---|---|---|---|
| General Info | |||
| Provider | Anthropic | Anthropic | Anthropic |
| Release Date | November 24, 2025 | July 24, 2026 | June 9, 2026 |
| Knowledge Cutoff | - | May 2026 | - |
| Context & Limits | |||
| Context Window | 200K | 1M Best Context Window | 1M Best Context Window |
| Pricing (per 1M tokens) | |||
| Input Pricing | $5 Best Input Pricing | $5 Best Input Pricing | $10 |
| Output Pricing | $25 Best Output Pricing | $25 Best Output Pricing | $50 |
| Modalities | |||
| Inputs | fileimagetext | textimage | textimagefile |
| Outputs | text | text | text |
| Benchmarks (0-100) | |||
| Intelligence Index | 35.6 | 63.1 Best Intelligence Index | 62.1 |
| Coding Index | - | 78.0 Best Coding Index | 76.5 |
| Agentic Index | - | 59.2 Best Agentic Index | 56.6 |
Graduate-level reasoning and expert Q&A evaluation.
Extremely difficult logical reasoning and knowledge.
Logical reasoning over long context windows.
Independent evaluation data provided by Artificial Analysis. To view the latest benchmarks and full details, visit their official site.
