Claude Opus 4.7
Claude Opus 4.7 by Anthropic: first Claude model with high-resolution vision, new tokenizer, adaptive thinking opt-in, real-time cybersecurity safeguards, and strict effort-level compliance.
Model Overview
Capabilities, design details, and architectural traits
Claude Opus 4.7 - First Claude Model With High-Resolution Vision and Real-Time Cyber Safeguards
Claude Opus 4.7 is Anthropic's most capable generally available model at launch, designed around sustained long-horizon autonomy - the ability to handle complex, multi-step tasks with rigor and consistency over extended runs. Its defining role in the Claude lineage is dual: it is the first Opus model with high-resolution image support, and the first Claude model deployed as a live testbed for Anthropic's real-time cybersecurity safeguards ahead of any Mythos-class public release.
What Is Structurally New in Opus 4.7
| Trait | What it means for Opus 4.7 specifically |
|---|---|
| First Claude model with high-resolution vision | Maximum image resolution increased to 2576px / 3.75MP, up from the previous 1568px / 1.15MP ceiling |
| New tokenizer | An updated tokenizer ships with 4.7 that may increase token counts by 1.0-1.35x compared to prior models, depending on content type |
| Adaptive thinking, opt-in | thinking: {type: "adaptive"} must be set explicitly; requests without it run without thinking - unlike Fable 5 and Mythos 5 where thinking is always on |
| Strict effort-level compliance | Documented to respect effort levels more strictly than Opus 4.6, especially at low and medium - it scopes work to what was asked rather than going beyond |
| Real-time cybersecurity safeguards - first deployment | Opus 4.7 is explicitly the first Claude model to carry automatic real-time detection and blocking of prohibited or high-risk cybersecurity requests, preceding any Mythos-class rollout |
| Thinking content omitted by default | Starting with Opus 4.7, thinking blocks appear in the stream but the thinking field is empty unless display: "summarized" is explicitly set |
The Cyber Safeguard Testbed Role
Anthropicexplicitly positioned Opus 4.7 as a deliberate stepping stone toward broader Mythos-class availability. During training, Anthropic experimented with differentially reducing its cybersecurity capabilities relative to Mythos Preview. Legitimate security professionals can apply to the Cyber Verification Program for exemption from these safeguards - a documented pathway unique to this model at launch.
Benchmark Performance
Independent evaluations · Artificial Analysis
Accuracy & Capability Details
Independent evaluation data provided by Artificial Analysis. To view the latest benchmarks and full details, visit their official site.
Compare Models Side-by-Side
Evaluate specifications, pricing, and independent benchmark indices
| Model Details | AnthropicHost Claude Opus 4.7 | ||
|---|---|---|---|
| General Info | |||
| Provider | Anthropic | Anthropic | Anthropic |
| Release Date | April 16, 2026 | June 9, 2026 | May 28, 2026 |
| Knowledge Cutoff | — | — | — |
| Context & Limits | |||
| Context Window | 1M | 1M | 1M |
| Pricing (per 1M tokens) | |||
| Input Pricing | $5 Best Input Pricing | $10 | $5 Best Input Pricing |
| Output Pricing | $25 Best Output Pricing | $50 | $25 Best Output Pricing |
| Modalities | |||
| Inputs | textimagefile | textimagefile | textimagefile |
| Outputs | text | text | text |
| Benchmarks (0-100) | |||
| Intelligence Index | 53.5 | 59.9 Best Intelligence Index | 55.7 |
| Coding Index | 73.6 | 76.5 Best Coding Index | 74.3 |
| Agentic Index | 44.4 | 52.8 Best Agentic Index | 47.2 |
GPQA Benchmark
Graduate-level reasoning and expert Q&A evaluation.
Humanity's Last Exam
Extremely difficult logical reasoning and knowledge.
IFBench Instruction
Evaluation of strict instruction following.
Benchmark scores are independent evaluations sourced from Artificial Analysis. Intelligence, Coding, and Agentic indices reflect composite performance ratings (0–100).