| Input per 1M tokens | $1 |
| Output per 1M tokens | $5 |
| Cache read per 1M tokens | $0.1 |
| Status | active |
| Source | official pricing page · verified 2026-09-07 |
Example costs
- One request (2,000 in / 500 out): $0.0045
- 100K such requests/month: $450
- 1M input + 1M output tokens: $6
Compare against other models →
How Claude Haiku 4.5 compares on rate
Claude Haiku 4.5 runs $1 input / $5 output per MTok — a combined $6 for 1M tokens each way. The cheapest active model in our table, Gemini 2.5 Flash-Lite at $0.1/$0.4 (combined $0.5), is about 92% cheaper on combined rate (verified 2026-09-07).
Cache reads on Claude Haiku 4.5 are priced at $0.1 per MTok — 0.1× the fresh input rate of $1. A workload where most requests share a long prefix pays far less than the base-rate estimate.
Per-token rate is not per-task cost — tokenizers differ between model families, so benchmark with your own payloads before switching.
What real volumes cost at Claude Haiku 4.5 rates
Straight rate-card math at the verified $1/$5 per-MTok prices, assuming a 75% input / 25% output token split — no caching or batch discounts applied.
| Volume | Split | Cost at Claude Haiku 4.5 rates |
|---|---|---|
| 100K tokens in a day | 75,000 in + 25,000 out | $0.2000 |
| That pace over a 30-day month (3M tokens) | 2.25M in + 750K out | $6.00 |
| 1M tokens in a month | 750,000 in + 250,000 out | $2.00 |
| Heavy month — 12M tokens | 9M in + 3M out | $24.00 |
Change the split or volume in the calculator — same verified rates, your numbers. Method + the corrections that make estimates real: how to estimate LLM API costs.
Frequently asked questions
What does Claude Haiku 4.5 cost per million tokens?
Claude Haiku 4.5 costs $1 per 1M input tokens and $5 per 1M output tokens on the Anthropic API, with cache reads at $0.1 per 1M tokens (verified 2026-09-07).
How much does 100K tokens a day cost on Claude Haiku 4.5?
At a 75% input / 25% output split, 100K tokens costs about $0.2000 at Claude Haiku 4.5 rates — about $6.00 over a 30-day month. That is straight rate-card math, before any caching or batch discounts.
Is there a cheaper model than Claude Haiku 4.5?
Yes — the cheapest active model we track is Gemini 2.5 Flash-Lite at $0.1/$0.4 per MTok, a combined $0.5 against Claude Haiku 4.5's $6 — about 92% less on combined rate (verified 2026-09-07). Tokenizers differ between families, so benchmark per-task cost with your own payloads.
How much do cache reads cost on Claude Haiku 4.5?
Cached input on Claude Haiku 4.5 is billed at $0.1 per 1M tokens — 0.1× the fresh input rate of $1. Workloads where most requests share a long prefix pay far below the base-rate estimate.