Anthropic API pricing
16 models tracked · headquartered in US · official pricing
Free tier: Unclear What Anthropic gives you free, in its own words
| Model | Input | Output | Context | |
|---|---|---|---|---|
| Claude Haiku 5.5 | $0.1 | $0.5 | 1M | reasoning vision |
| Claude Haiku 4.5 | $1 | $5 | 200K | reasoning vision |
| Claude Sonnet 5 | $2 | $10 | 1M | reasoning vision |
| Claude Sonnet 5.5 | $2 | $10 | 1M | reasoning vision |
| Claude Sonnet 4.5 legacy | $3 | $15 | 200K | reasoning vision |
| Claude Sonnet 4.6 | $3 | $15 | 1M | reasoning vision |
| Claude Opus 5.5 | $4 | $20 | 1M | reasoning vision |
| Claude Opus 4.5 legacy | $5 | $25 | 200K | reasoning vision |
| Claude Opus 4.6 | $5 | $25 | 1M | reasoning vision |
| Claude Opus 4.7 | $5 | $25 | 1M | reasoning vision |
| Claude Opus 4.8 | $5 | $25 | 1M | reasoning vision |
| Claude Opus 5 | $5 | $25 | 1M | reasoning vision |
| Claude Fable 5 2 | $8 | $40 | 1M | reasoning vision |
| Claude Fable 5 | $10 | $50 | 1M | reasoning vision |
| Claude Fable 5.1 | $10 | $50 | 1M | reasoning vision |
| Claude Opus 4.1 retired | $15 | $75 | 200K | reasoning vision |
Compare Anthropic’s models on your own workload
How Anthropic prices its models
Output prices across Anthropic's 13 current models run from $0.5 per 1M tokens (Claude Haiku 5.5) to $50 (Claude Fable 5.1), a 100x spread. Picking the smallest model that does the job is usually the biggest saving available.
Discounts: caching and batch
13 of 13 current models list a cached-input rate, from 90% off the fresh input rate on Claude Haiku 5.5. 8 list a batch discount for work that can wait.
Recorded price changes
- 2026-09-01: Claude Sonnet 5 input $3 to $2
- 2026-09-01: Claude Sonnet 5 output $15 to $10
- 2026-08-22: Claude Sonnet 5 input $2 to $3
- 2026-08-22: Claude Sonnet 5 output $10 to $15
- 2026-08-16: Claude Sonnet 5 input $3 to $2
- 2026-08-16: Claude Sonnet 5 output $15 to $10
Where these prices come from
Every rate above is read from Anthropic's own pricing page (linked here) and re-checked every morning. A disputed figure is marked rather than hidden; see how PromptSpend checks prices.
Common questions
How much do Anthropic API models cost?
PromptSpend tracks 16 Anthropic models. The cheapest on a blended rate is Claude Haiku 5.5, at $0.1 per 1M input tokens and $0.5 per 1M output tokens; the most expensive is Claude Opus 4.1, at $15 in and $75 out.
Does Anthropic have a free API tier?
PromptSpend's verdict is "Unclear". Its free-tier page gives Anthropic's own wording for each claim, with a link to the page it came from and the date it was read.
Where do the Anthropic prices come from?
From Anthropic's own pricing page, https://platform.claude.com/docs/en/about-claude/pricing, with each model's check date shown on its page. Prices change without notice, and regional premiums, priority tiers and negotiated discounts are not included.