Anthropic · flagship
Claude Opus 5: API pricing 2026
Anthropic's current flagship for complex agentic coding and enterprise work (per claude.com/pricing), at $5/$25 — the same headline rate as the Opus 4.x line it succeeds. Above it sits the Fable 5.1 frontier tier ($10/$50); below it, Sonnet 5 handles most production automation at a fraction of the cost. Anthropic's larger models can tokenize the same source text into more tokens than the smaller tiers, so validate real cost-per-task against your own prompts before committing.
Input $/M tokens
$5.00
Output $/M tokens
$25.00
Cost / task
$0.0400
3k in + 1k out
Tier
flagship
Anthropic
Cost at common task volumes
| 1,000 tasks / yr | $40.00 |
| 10,000 tasks / yr | $400 |
| 100,000 tasks / yr | $4,000 |
| 1,000,000 tasks / yr | $40,000 |
Token shape: 3,000 input + 1,000 output per task. No prompt caching. Real cost varies with your actual prompt shape.
Source
↗ https://claude.com/pricingVerified 2026-09-15.
Compare with other LLMs