Claude Haiku 5.5 API pricing
Anthropic's smallest and fastest model. Claude Haiku 5.5 launched on 7 October 2026 at $0.10 / $0.50 per 1M tokens for prompts up to 100K tokens, 90% less than Haiku 4.5, with a 1M-token context window.
Claude Haiku 5.5 prices
Joint 3rd cheapest of 12 current small and fast models (blended 3:1). Cheapest: Ministral 3 3B, $0.10 / $0.10.
Capability score 160, 9th of 29 current models scored on the Epoch Capabilities Index (range 158–163).
No cheaper current model scores level or higher on a coding agent ($34.59 a month).
Source: Epoch Capabilities Index by Epoch AI, CC BY 4.0, read 11 Nov 2026. What the score means
What it costs in practice
Claude Haiku 5.5 at list prices, no batch discount.
| 1,000 chat replies1,500 tokens in and 300 out each | $0.30 |
|---|---|
| Summarising 100 long reports20,000 tokens in and 800 out each | $0.24 |
| A month of a coding agent400 tasks, 85% of input cached | $21.63 |
The coding-agent example assumes 85% of input is served from cache, typical for agents that resend the same context each step, and prompts under 100K tokens; longer prompts cost more. Work out AI agent cost on Claude Haiku 5.5
Details
claude-haiku-5-5- Prompts over 100K tokens cost $0.50 / $2.50 per 1M, with cache reads at $0.05. The whole request is billed at the higher rate.
- Cache writes cost $0.125 per 1M tokens ($0.625 for prompts over 100K).
- 90% cheaper per token than Claude Haiku 4.5. Anthropic says it costs about 75% less to run on average.
Checked 11 Nov 2026 against Anthropic pricing and Anthropic, 7 Oct 2026 .
Estimate your monthly cost
List prices, no batch discount
| Model | Input / output per 1M | Per month |
|---|---|---|
| Claude Haiku 5.5 | $0.10 / $0.50 | $2.00 |
| Ministral 3 3BCheaper alternative · Mistral AI | $0.10 / $0.10 | $1.20 |
| GPT-6 LunaSimilar price · OpenAI | $0.10 / $0.50 | $2.00 |
Type amounts like 500K, 25M or 2B.
Price Claude Haiku 5.5 in: Token cost calculator · AI agent cost · Token counter
Cheaper alternative
Ministral 3 3B (Mistral AI) costs $0.10 / $0.10 per 1M tokens, 50% less than Claude Haiku 5.5 at a 3:1 input-to-output mix. It’s the cheapest current small and fast model from another family. Cheaper doesn’t mean as good: test it on your own tasks.
Compare Claude Haiku 5.5 with Ministral 3 3B
Similar price from another provider: GPT-6 Luna, $0.10 / $0.50. Compare
Compare Claude Haiku 5.5 with
Side by side: prices, cached and batch rates, and the cost of the same workload.
Other Claude Haiku versions
Prompts over 100K tokens cost $0.50 / $2.50. Haiku 4.5 is still served at $1 / $5.
| Model | Input / 1M | Output / 1M | Cached input | Context | Released |
|---|---|---|---|---|---|
| $1.00 | $5.00 | $0.10 | 200K | 15 Oct 2025 |
Claude Haiku 4.5 details
Claude Haiku 4.5
Previous$1.00 / $5.00 per 1M tokenscached $0.10 · batch 50% off · 200K context · 64K max output
Newer option: Claude Haiku 5.5 at $0.10 / $0.50. Checked 11 Nov 2026 against Anthropic pricing .
Compare Claude Haiku 5.5
Price history
All changes- ▼ 90%Claude Haiku 5.5 replaced Haiku 4.5 at $0.10 / $0.50 per 1M tokens for prompts up to 100K, 90% cheaper per token.
$1 / $5→ $0.10 / $0.50 Source
Every dated price: price records · CSV. Cite as: AI Token Price, Claude Haiku pricing, https://aitokenprice.com/pricing/claude-haiku#price-history
Anthropic pricing rules
- Batch requests cost 50% less.
- Cache reads are billed at a per-model rate, usually a tenth of the input price or less. Cache writes cost 1.25× input for the 5-minute cache and 2× for the 1-hour cache.
- The full 1M-token context window is billed at standard rates on current models.
Questions
How much does Claude Haiku 5.5 cost?
As of 11 Nov 2026, Claude Haiku 5.5 costs $0.10 per million input tokens and $0.50 per million output tokens, and $0.01 per million cached input tokens. Batch requests cost 50% less. Prices checked 11 Nov 2026.
What is Claude Haiku 5.5's context window?
1M tokens, with up to 128K tokens of output.
Does Claude Haiku 5.5 support prompt caching?
Yes. Repeated input read from the cache costs $0.01 per million tokens, 90% less than normal input.