Claude Opus 4.7
Anthropic's latest Opus model, released April 2026 and (#3 on LMRank). Excels in complex reasoning, code generation, and agentic tasks with a 1M context window. Priced at $5/MTok input, $25/MTok output. Now superseded by Opus 4.8 (May 2026).
Specifications
| Lab | Anthropic |
|---|---|
| Context window | 1M tokens |
| Input price | $5.00/1M |
| Output price | $25.00/1M |
| Release | Apr 2026 |
Strengths
- Top-tier complex reasoning and code generation
- Strong gains on SWE-bench Pro (+10.9%)
- MCP-Atlas improvement of +14.6%
- Supports adaptive-only thinking with 'xhigh' effort
Weaknesses
- Slower output speed (52 t/s)
- High time-to-first-token (18.38s)
- Token inflation from new tokenizer
- Superseded by Opus 4.8 shortly after release
Best for
- Complex reasoning tasks
- Code generation and debugging
- Agentic and autonomous workflows
- Long-context research analysis
In Depth: Claude Opus 4.7
Benchmark Performance
Claude Opus 4.7 , ranking #3 overall. It leads peers in agentic and code benchmarks, though its reign was brief before Opus 4.8 arrived.
Anthropic reports Opus 4.7 improved SWE-bench Pro by +10.9% and MCP-Atlas by +14.6% over Opus 4.6. These are vendor-reported gains. The model also shows strong results on complex reasoning tasks, though independent verification of all claims is pending.
Pricing & Value
Priced at $5.00 per million tokens input and $25.00 per million tokens output a premium tier reflecting its advanced capabilities.
At $5/$25 per MTok, Opus 4.7 sits in the same price band as Opus 4.6 and 4.8, well above open-weight alternatives like DeepSeek V4 Pro. The premium buys agentic and coding strength; for bulk workloads, cheaper models cover most tasks.
Who Should Use This
Best for developers needing cutting-edge agentic performance. Also suits researchers tackling long-context reasoning. Not ideal for latency-sensitive applications.
- Agentic engineers: unmatched autonomy, but slow token generation
- Code specialists: strong SWE-bench gains, watch for token inflation
- Deep researchers: 1M context, high cost per call
- Anti-persona: real-time apps needing sub-second responses
Release & Version History
Released April 16, 2026, Claude Opus 4.7 succeeded Opus 4.6 but was itself superseded just six weeks later by Opus 4.8 on May 28, 2026.
Opus 4.7 arrived as Anthropic's most powerful model, offering a 1M context window and new 'xhigh' effort mode. It was available via Anthropic API, AWS Bedrock, GCP Vertex AI, and Microsoft Foundry. Despite strong benchmark gains, Anthropic quickly iterated to Opus 4.8, making 4.7 a short-lived but impactful release in the Opus lineage.