Claude Opus 4.7
Anthropic's flagship model, released 16 April 2026 — a single-version jump that took SWE-bench Verified past 87% and ahead of Gemini 3.1.
Claude Opus 4.7 specs at a glance
| Maker | Anthropic |
| Released | 16 April 2026 |
| Context window | 1,000,000 tokens |
| Max output | Up to 128,000 tokens |
| API pricing | $5 / $25 per 1M input / output tokens |
| Cost savings | Up to 90% with prompt caching · 50% with batch processing |
| Reasoning control | New 'xhigh' effort level · task budgets (public beta) |
| Best at | Coding, agentic work, long-context reasoning, vision |
What's new in Claude Opus 4.7
Claude Opus 4.7 is a point release that behaves like a major one. Anthropic positions it as its most capable generally available model, performing at the frontier across coding, agentic tasks and knowledge work — and the benchmark jumps back that up.
Two developer-facing additions stand out. The first is a new 'xhigh' effort level, slotted between the existing 'high' and 'max' settings, giving finer control over the reasoning-versus-latency trade-off. The second is task budgets, in public beta on the Claude Platform, which let teams cap how much compute a task is allowed to consume.
Claude Opus 4.7 benchmarks
The headline is the size of the gains in a single version bump.
| Benchmark | Opus 4.6 | Opus 4.7 |
|---|---|---|
| SWE-bench Verified | 80.8% | 87.6% |
| SWE-bench Pro | 53.4% | 64.3% |
Key takeaway
SWE-bench Pro climbed 10.9 points and SWE-bench Verified nearly 7 points in one release — putting Opus 4.7 ahead of Gemini 3.1 Pro (80.6% Verified) and giving it the most consistent long-context performance of any model tested.
Pricing and the tokenizer caveat
Opus 4.7 keeps the same rate card as Opus 4.6: $5 per 1M input tokens and $25 per 1M output tokens, with up to 90% savings from prompt caching and 50% from batch processing.
But there is a catch worth knowing. Opus 4.7 ships with a new tokenizer that can produce up to 35% more tokens for the same input text. Because you are billed per token, your real cost per request can rise even though the per-token price did not move. When budgeting, compare end-to-end request costs, not just the headline rate.
Vision improvements
Opus 4.7 substantially upgrades vision. It now supports high-resolution images up to 2,576 pixels on the long edge — roughly 3.75 megapixels, over 3x the resolution prior Claude models accepted. In practice that unlocks dense screenshot reading, complex diagram extraction and pixel-accurate UI references that earlier versions struggled with.
Who should use Claude Opus 4.7
- Software engineering teams — its 87.6% SWE-bench Verified score is among the highest of any model.
- Long-context workloads — Anthropic reports the most consistent long-context performance of any model tested.
- Vision-heavy tasks — dense screenshots, diagrams and high-resolution image analysis.
- Cost-aware teams that can cache — prompt caching cuts up to 90% off repeated-context workloads.
Watch the tokenizer change if you are migrating from Opus 4.6, and consider cheaper open models such as DeepSeek V4 or Kimi K2.6 for high-volume coding where the absolute frontier is not required.
How Claude Opus 4.7 compares
- GPT-5.5 ("Spud") overview →
- Gemini 3.1 Ultra overview →
- DeepSeek V4 — the open-weights challenger →
- GPT-5.5 vs Gemini 3.1 Ultra →
Frequently asked questions
When was Claude Opus 4.7 released?
Anthropic released Claude Opus 4.7 on 16 April 2026. It is Anthropic's most capable generally available model.
How much does Claude Opus 4.7 cost?
Claude Opus 4.7 costs $5 per million input tokens and $25 per million output tokens — the same rate card as Opus 4.6. Prompt caching can cut costs by up to 90% and batch processing by 50%.
What is Claude Opus 4.7's context window?
Claude Opus 4.7 has a 1 million token context window and can produce up to 128,000 output tokens.
How good is Claude Opus 4.7 at coding?
Claude Opus 4.7 scores 87.6% on SWE-bench Verified — up from 80.8% on Opus 4.6 — and 64.3% on SWE-bench Pro, up from 53.4%. The SWE-bench Verified score puts it ahead of Gemini 3.1 Pro.
Did Claude Opus 4.7 get more expensive?
The rate card is unchanged from Opus 4.6, but Opus 4.7 ships with a new tokenizer that can produce up to 35% more tokens for the same input text. Your real bill per request can rise even though the per-token price did not.
What is the 'xhigh' effort level?
'xhigh' is a new reasoning effort level introduced in Opus 4.7, sitting between 'high' and 'max'. It gives developers finer control over the trade-off between reasoning depth and response latency.