Anthropic released Claude Opus 4 as a major upgrade over Opus 3.5. Here’s what changed and whether it matters for your workflow.
Key changes
| Opus 3.5 | Opus 4 | |
|---|---|---|
| Context window | 200K | 200K |
| SWE-bench score | ~49% | ~76.8% |
| Agentic capabilities | Limited | Strong (tool use, multi-step) |
| Input price | $15 / 1M tokens | $15 / 1M tokens |
| Output price | $75 / 1M tokens | $75 / 1M tokens |
What improved
Coding ability — The biggest jump. Opus 4 nearly doubled the SWE-bench score, going from ~49% to ~76.8%. It handles complex multi-file changes, understands project structure better, and produces more complete solutions.
Agentic workflows — Opus 4 is significantly better at using tools, making multi-step plans, and executing complex tasks autonomously. This matters if you’re building AI agents or using tools like Kiro.
Instruction following — Opus 4 is more precise at following detailed instructions and less likely to go off-track on complex prompts.
What stayed the same
- Context window (still 200K)
- Pricing (same input/output costs)
- Multimodal support (text + images)
Should you upgrade?
Yes, if you use Claude for coding, agentic tasks, or complex analysis. The improvement is substantial.
Doesn’t matter if you mainly use Claude for simple chat, summarization, or writing. The difference is less noticeable for basic tasks.
Since pricing is identical, there’s no reason not to use Opus 4 if you have access.
See our full AI Model Comparison for all models side by side.