Model pricing · Anthropic
Claude Haiku 5.5 pricing
Claude Haiku 5.5 costs $0.1 per million input tokens and $0.5 per million output tokens at list price. Best for: fast, cheap subtasks.
| Input | $0.1 / 1M tokens |
|---|---|
| Output | $0.5 / 1M tokens |
| Cached input (read) | $0.01 / 1M tokens |
| Cache write | $0.125 / 1M tokens |
| Context window | Not published |
Base tier shown: prompts up to 100K tokens. A request whose prompt exceeds 100K bills $0.50/$2.50; cache reads and writes count toward prompt length. Context window is not stated on the pricing page.
Source: Anthropic pricing, cross-checked against models.dev on 2026-11-13. Your provider invoice is authoritative.
What it costs in practice
| Workload | List price | If input is cached |
|---|---|---|
| One request: 10,000 tokens in, 1,000 out | $0.0015 | $0.0006 |
| A month of that request, 100 times a day | $4.50 | $1.80 |
| A 3,000-token instruction file on 100 requests a day, for a month | $0.900 | $0.090 |
The last row is the always-on cost of an instruction file like CLAUDE.md: paid on every request, whether the task needs it or not. Check how much of yours is waste.
Similarly priced models
Gemini 2.5 Flash-Lite
$0.1 in · $0.4 out · Google
Llama 4 Scout
$0.1 in · $0.3 out · Meta (via OpenRouter)
Amazon Nova Lite
$0.06 in · $0.24 out · Amazon (Bedrock)
DeepSeek V4 Flash
$0.14 in · $0.28 out · DeepSeek
Compare every model side by side or price your own workload.