API pricing and cost
Claude Haiku 5.5 pricing
Claude Haiku 5.5 official API pricing is $0.10 per million input tokens and $0.50 per million output tokens. Third-party provider prices are compared below.
Covers the 2026-10-07 Claude Haiku 5.5 release. Supersedes Haiku 4.5.
$0.10
Official input / 1M
Standard tier
$0.50
Official output / 1M
Standard tier
$0.115
Blended / 1M
4:1 input/output, 90% input cached
$11.52
Example monthly cost
10K requests, 8K input + 2K output/request, 90% input cached
Token charges only; tool calls, taxes, and other non-token fees are excluded.
Official API pricing
USD / 1M text tokens · Standard · Effective from 2026-10-07 · Last verified 2026-10-07
- InputStandard text tier, per million tokens
- $0.10
- Cached inputInput served from the prompt cache
- $0.01
- Cache writeCharged once when a prompt enters the cache
- $0.125
- OutputStandard text tier, per million tokens
- $0.50
Billing conditions and workload budgets
Pricing review:
Anthropic lists input/output/cache-read rates of $0.10/$0.50/$0.01 per million tokens at up to 100K prompt tokens, and $0.50/$2.50/$0.05 above 100K. Cache writes are additional: $0.125 or $0.625 for the listed write tier. Regional processing and tools can add charges. Treat the default calculator as an entry-tier token estimate.
The default workloads fit the entry prompt tier. The calculator applies the published higher context-length band when input per request exceeds 100K. Include system instructions, tool definitions and conversation history in that prompt size; add cache-write charges separately.
API providers (4)
Published USD text API rates across hosting providers for Claude Haiku 5.5. Compare listed rates, effective dates, and service-tier notes for the same model.
| Provider | Input / 1M | Cached / 1M | Output / 1M | Additional tiers & notes |
|---|---|---|---|---|
| AmazonEffective from 2026-11-14 | $0.10 | $0.01 | $0.50 | The displayed row is the base standard text-token rate. Additional standard rate bands apply. |
| AnthropicOfficialEffective from 2026-10-07 | $0.10 | $0.01 | $0.50 | The displayed row is the base standard text-token rate. Additional standard rate bands apply. Batch: $0.05/$0.25 per 1M tokens. |
| Google Cloud Vertex AIEffective from 2026-11-14 | $0.10 | $0.01 | $0.50 | The displayed row is the base standard text-token rate. Additional standard rate bands apply. Batch: $0.05/$0.25, $0.005 cached per 1M tokens. |
| MicrosoftEffective from 2026-11-14 | $0.10 | $0.01 | $0.50 | The displayed row is the base standard text-token rate. Additional standard rate bands apply. Batch: $0.05/$0.25 per 1M tokens. |
What Claude Haiku 5.5 costs to run
Four workloads costed at official standard rates. Cached-input share is an estimation assumption, not a provider guarantee; when no cached rate is published, all input is billed at the uncached rate. Non-token fees are excluded.
| Workload | Per request | Monthly tokens | Input | Output | Per month |
|---|---|---|---|---|---|
| Chatbotshort turns | 1,500 in · 500 out× 100,000 requests | 150M in · 50M out | $15.00 | $25.00 | $40.00$0.0004 / request |
| Assistant8K context · 90% cached | 8,000 in · 2,000 out× 10,000 requests | 80M in · 20M out | $1.52 | $10.00 | $11.52$0.001152 / request |
| RAGlong input, cached · 50% cached | 24,000 in · 800 out× 20,000 requests | 480M in · 16M out | $26.40 | $8.00 | $34.40$0.00172 / request |
| Coding agentlong output · 70% cached | 15,000 in · 4,000 out× 20,000 requests | 300M in · 80M out | $11.10 | $40.00 | $51.10$0.002555 / request |
Monthly cost
$11.52
on Claude Haiku 5.5
Per request
$0.001152
80M in · 20M out a month
Split
$1.52 + $10.00
input + output
The same workload costs $11.09 on Muse Ember, the cheapest of the comparable models below.
- Muse EmberFacebook AI Research Lab · $0.111 / 1M blended$11.09
- GPT-6 LunaOpenAI · $0.115 / 1M blended$11.52
- Phi-4-mini-reasoningMicrosoft · $0.12 / 1M blended$12.00
- Claude Sonnet 5.5Anthropic · $2.23 / 1M blended$223.20
- Claude Fable 5.1Anthropic · $10.98 / 1M blended$1,098.00
- GPT-6 AstraOpenAI · $11.52 / 1M blended$1,152.00
Claude Haiku 5.5 against comparable models
Benchmark-overlap peers and price-comparable same-type models, costed on the assistant workload above. All figures use the same workload and caching rule; lower price does not imply equivalent capability or service terms.
| Model | Basis | Blended / 1M | Per month | vs Claude Haiku 5.5 |
|---|---|---|---|---|
| Muse EmberFacebook AI Research Lab | Lower-cost option | $0.111 | $11.09 | 3.8%cheaper |
| Claude Haiku 5.5Anthropic · this model | This model | $0.115 | $11.52 | — |
| GPT-6 LunaOpenAI | Similar price | $0.115 | $11.52 | Same |
| Phi-4-mini-reasoningMicrosoft | Similar price | $0.12 | $12.00 | 4.2%dearer |
| Claude Sonnet 5.5Anthropic | Shared benchmark coverage | $2.23 | $223.20 | 1,837%dearer · 19.4× |
| Claude Fable 5.1Anthropic | Shared benchmark coverage | $10.98 | $1,098.00 | 9,431%dearer · 95.3× |
| GPT-6 AstraOpenAI | Shared benchmark coverage | $11.52 | $1,152.00 | 9,900%dearer · 100× |
Catalog price position: #35 cheapest of 269 active priced models in the LLMLearner catalog (cheaper than 87% at the 4:1 mix with 90% of input assumed cached where published; official headline rates).
Claude Haiku 5.5 pricing questions
- How much does Claude Haiku 5.5 cost per million tokens?
- Claude Haiku 5.5 official API pricing is $0.10 per million input tokens and $0.50 per million output tokens. Cached input is billed at $0.01 per million tokens. At the default 4:1 input-to-output mix with 90% of input assumed cached, that is $0.115 per million blended tokens.
- What does Claude Haiku 5.5 cost per month, and what is included?
- On the example Assistant workload (10,000 requests/month, 8,000 input and 2,000 output tokens/request, 90% input cached), Claude Haiku 5.5 costs about $11.52/month ($0.001152/request). This includes token charges at the official rate only; tool calls, off-peak or batch tiers, and taxes are excluded.
- Is there a cheaper alternative to Claude Haiku 5.5?
- Among the comparable models shown on this page, Muse Ember is cheaper at $0.111 per million blended tokens vs $0.115 for Claude Haiku 5.5. Compare benchmarks before switching, as a lower rate does not guarantee equivalent quality or service terms. Across all 269 active priced models in the LLMLearner catalog, Claude Haiku 5.5 ranks 35th cheapest (cheaper than 87% at the default 4:1 mix).
- Does Claude Haiku 5.5 charge for cached input?
- Yes, at a discount: $0.01 per million cached input tokens against $0.10 uncached, with cache writes billed at $0.125 per million.