Claude Opus 5 vs Opus 4.8 vs Sonnet 5: Which? (2026)
The Short Answer
Anthropic’s July 2026 lineup splits by job: Sonnet 5 for everyday work, Opus 5 for the hardest tasks, Opus 4.8 now superseded. Opus 5 (launched July 24) is the new default Opus at the same $5/$25 as 4.8 — so there’s no reason to stay on 4.8.
The Lineup
| Opus 5 | Opus 4.8 | Sonnet 5 | |
|---|---|---|---|
| Launched | Jul 24, 2026 | prior | Jun 30, 2026 |
| Price ($/MTok) | $5 / $25 | $5 / $25 | $2 / $10 intro* |
| SWE-bench Verified | 96.0% | lower | high |
| Context | 1M | 1M | large |
| Best for | Hardest tasks | (superseded) | Everyday work |
*Sonnet 5 intro pricing runs through Aug 31, 2026, then rises to $3/$15.
Opus 5 vs Opus 4.8
Opus 5 replaces Opus 4.8 at the same price. It’s now the default Opus, posts a higher SWE-bench Verified (96.0%), keeps the 1M context and 128K max output, and lowers the cache minimum to 512 tokens (cheaper caching for shorter prompts). If you’re on Opus 4.8, moving to Opus 5 is a free upgrade.
Opus 5 vs Sonnet 5
This is the real decision. Sonnet 5 costs less than half ($2/$10 intro vs $5/$25) and handles the majority of coding and chat well. Reserve Opus 5 for:
- Hard multi-repo refactors where a wrong edit is expensive
- Accuracy-critical agentic loops
- Tasks where its 96% SWE-bench Verified lead actually shows up
For everything else, Sonnet 5 is the value pick. Note the intro pricing ends August 31, 2026 — after that Sonnet 5 is $3/$15, still well under Opus.
Which Claude Should You Use?
- Everyday coding / chat → Sonnet 5
- Hardest refactors / accuracy-critical agents → Opus 5
- Cheapest / simple tasks → Haiku 4.5 ($1/$5)
- Already on Opus 4.8 → switch to Opus 5 (same price, better)
Sources
- Anthropic — Claude Opus 5: anthropic.com/news/claude-opus-5
- Anthropic — Claude Sonnet 5: anthropic.com/news/claude-sonnet-5
- Anthropic — pricing: anthropic.com/pricing