# Claude Opus 5 review: 10/10

URL: https://ongchummo.vn/en/review-ai-tool/claude-opus-5

Category: Frontier LLM · Pricing: Paid
Tool website: https://www.anthropic.com/claude/opus

## Verdict

The current top model in the table, strongest at agentic coding and long-running, multi-step tasks.

## Scores by task (out of 10)

- Video: 8.6
- Image: 9.4
- 3D: 8.2
- Chat: 9.8
- Code: 10
- Web: 10

## Strengths

- Leads the AA Intelligence Index v4.1 with 63 points
- #1 on Agentic Index at 55.3 and WebDev Arena at 1702.9
- 1M context, can read an entire large repo in one pass

## Weaknesses

- $5/$25 per million tokens, pricier than the balanced tier
- You have to pick the right reasoning level (low to max) to keep costs optimized
- The max level burns a lot of reasoning tokens

## Full review

### Ranking source
Updated August 9, 2026 from the Artificial Analysis Intelligence Index (v4.1 scale), cross-checked against each provider's release notes and official pages. Scores on the v4.1 scale can't be compared directly with earlier rankings.

### When to use it
Use it for the hardest jobs: refactoring large repos, multi-step agents, and long tasks that need reliability. Terminal-Bench 2.1 reaches 86.7%, ARC-AGI-3 30%, GDPval-A v2 1858.

### Notes
AA Intelligence Index: 63 (max/xhigh); context 1M; released July 24, 2026; price $5/$25. Prioritize it when output quality matters more than price per token.
