Claude Opus 4.8
anthropic / claude-opus-4-8 / official pricing
Context 1000k · Max output 128k · Released 2026-05-28 · Last verified 2026-06-12
Standard pricing (per 1M tokens)
| Tier | Input | Output |
|---|---|---|
| Standard | $5.00 | $25.00 |
Prompt caching
| Type | Price / 1M |
|---|---|
| Cache type | explicit |
| Cache read (hit) | $0.50 |
| Cache write (5-min TTL) | $6.25 |
| Cache write (1-hour TTL) | $10.00 |
| Minimum cacheable tokens | 1024 |
Want to model your real cache savings? Try the cache calculator with this model selected.
Batch API
- Input discount50%
- Output discount50%
- Stacks with cachingyes
See the batch decision tool for whether this fits your workload.
Capabilities
caching batch tool use vision reasoning extended thinking
Notes & extras
- Web search: $10.00 per 1,000 searches
- Reasoning tokens: billed as output
- Fast mode (research preview) runs this model at $10/$50 per 1M with faster output. Uses the tokenizer introduced with Opus 4.7: the same text produces roughly 30% more tokens than older Claude models, so per-token estimates based on older tokenizers undercount.
Sources
Found a price wrong? Report it