## Anthropic (Claude API) pricing

Pricing page: https://claude.com/pricing#api. Verified 2026-10-22.
Lock-in: Low to medium: the Messages API is simple to swap, but prompts, tool definitions and output quality need re-testing on another model.

### Claude Haiku 5.5 (`haiku-5-5`): $0/month base
- Input tokens (per million): 0 included; then: $0.1 per 1,000
- Output tokens (per million): 0 included; then: $0.5 per 1,000
- Includes: commercial_use, compliance.hipaa, compliance.gdpr_dpa, ai.text, ai.vision
- Not yet priced: prompts over 100K tokens ($0.50 input, $2.50 output per million), prompt caching (cache reads cost 0.1x input on this model), Batch API (50% off input and output), US-only inference (1.1x), HIPAA: the BAA covers the first-party API only with zero data retention, arranged through sales, and not web search, batches, prompt caching or Files API uploads
- Note: $0.10 input and $0.50 output per million tokens for prompts up to 100K tokens; $0.50 and $2.50 for prompts over 100K tokens. Images cost ceil(width/28) x ceil(height/28) visual tokens. Cheapest model: compare quality in a spike before relying on it.

### Claude Sonnet 5.5 (`sonnet-5-5`): $0/month base
- Input tokens (per million): 0 included; then: $2 per 1,000
- Output tokens (per million): 0 included; then: $10 per 1,000
- Includes: commercial_use, compliance.hipaa, compliance.gdpr_dpa, ai.text, ai.vision
- Not yet priced: prompt caching (cache reads cost 0.05x input on this model), Batch API (50% off input and output), US-only inference (1.1x), HIPAA: the BAA covers the first-party API only with zero data retention, arranged through sales, and not web search, batches, prompt caching or Files API uploads
- Note: $2 input and $10 output per million tokens.

### Claude Opus 5.5 (`opus-5-5`): $0/month base
- Input tokens (per million): 0 included; then: $4 per 1,000
- Output tokens (per million): 0 included; then: $20 per 1,000
- Includes: commercial_use, compliance.hipaa, compliance.gdpr_dpa, ai.text, ai.vision
- Not yet priced: prompt caching (cache reads cost 0.05x input on this model), Batch API (50% off input and output), US-only inference (1.1x), fast mode ($8 input, $40 output per million), HIPAA: the BAA covers the first-party API only with zero data retention, arranged through sales, and not web search, batches, prompt caching or Files API uploads
- Note: $4 input and $20 output per million tokens. Fast mode costs $8 and $40.

### Claude Fable 5.1 (`fable-5-1`): $0/month base
- Input tokens (per million): 0 included; then: $10 per 1,000
- Output tokens (per million): 0 included; then: $50 per 1,000
- Includes: commercial_use, compliance.hipaa, compliance.gdpr_dpa, ai.text, ai.vision
- Not yet priced: prompt caching (cache reads cost 0.025x input on this model), Batch API (50% off input and output), US-only inference (1.1x), HIPAA: the BAA covers the first-party API only with zero data retention, arranged through sales, and not web search, batches, prompt caching or Files API uploads
- Note: $10 input and $50 output per million tokens.

**Sources**
- Anthropic (Claude API): verified 2026-10-22 (https://claude.com/pricing#api)

Confidence: high.

The full result is in the structured content.