Claude Haiku 4.5
How much does Claude Haiku 4.5 cost?
Claude Haiku 4.5 costs $1.00 per million input tokens and $5.00 per million output tokens on Anthropic's own API. Input read from the prompt cache costs $0.10 per million tokens. Checked against Anthropic's pricing page on October 12, 2026.
- Input Price
- $1.00
- per 1M tokens
- Output Price
- $5.00
- per 1M tokens
- Context Window
- 200K
- tokens
- Speed Score
- 97
- out of 100
Benchmarks
Avg 84/100- Coding83/100
- Reasoning82/100
- Extraction91/100
- Creative82/100
- Vision80/100
Pricing
Input
Prompt / context tokens
$1.00/ 1M
Output
Generated tokens
$5.00/ 1M
Batch API
Async processing discount
$0.50 / $2.50
50% off standard price
List price from the Anthropic pricing page, verified October 12, 2026.
Its tokenizer produces about 10% more tokens than OpenAI's for the same text, so the real cost runs higher than the list price suggests. This is an estimate: the tokenizer has not been measured yet.
Check the sourceBest For
Claude Haiku 4.5 in context
On a typical 70% input / 30% output mix, Claude Haiku 4.5 blends to $2.20 per million tokens. That puts it above 39% of the 46 models tracked here, and 15 other models in the same budget tier undercut it. Tier is a capability label, not a price band — the spread inside one tier is often wider than the gap between tiers, which is why picking by tier alone tends to overspend.
Input runs $1.00 per million and output $5.00 — a 5.0x ratio. That ratio matters more than either number on its own: a summarisation or classification workload reads far more than it writes and will track the input price, while a code-generation or long-form writing workload inverts that and will track the output price. Compare models on the ratio your own traffic actually has, not on the input price alone.
The benchmark profile is uneven rather than flat: data extraction at 91/100 against vision at 80/100, a 11-point spread. An average hides that. If your workload sits on the strong end this model punches above its price; if it sits on the weak end, a cheaper model with a flatter profile will serve you better.
The context window is 200K tokens and the throughput score is 97/100. Context only earns its keep if you fill it, and filling it is also what makes a request expensive — a large window is an option, not a discount. For a budget-tier model, judge the speed score against what you are doing: it decides everything for interactive chat and almost nothing for overnight batch work.
Worth checking before you commit: GPT-6 Luna scores at least as well on combined coding and reasoning and blends to $0.22 per million against Claude Haiku 4.5's $2.20. That does not make Claude Haiku 4.5 the wrong choice — context window, latency, vendor lock-in, and specific capabilities all sit outside a benchmark score — but it does mean the price difference needs a reason behind it.
The closest comparison is DeepSeek V3.2 from DeepSeek, at $0.32 per million against $2.20. 7 models in this tier score higher on combined coding and reasoning, which is the useful framing: the question is rarely whether a model is good, it is whether it is the cheapest thing that is good enough for the specific work you are sending it.
What it costs in production
At 10M + 2M tokens a month — a realistic mid-size production workload — Claude Haiku 4.5 runs about $20.00. Routing the latency-tolerant share through the batch API at 50% off would bring that to roughly $10.00, which is usually a bigger saving than switching models.
What you actually pay
The list price is only the start. Prompt caching, batch processing and long prompts change the bill, and each provider sets them differently. Prices per 1M tokens.
| Pricing term | Claude Haiku 4.5 |
|---|---|
| List price (input / output) | $1.00 / $5.00 |
| Cached input | $0.1010% of the input price |
| Batch API | 50% off input and output |
| Long prompts | No surcharge across the 200K window |
| Input with 80% cache hits | $0.28 |
The last row assumes 80% of input tokens are read from the cache, as in a chat app that re-sends a long system prompt or RAG over the same documents. Cache-write and storage charges are not included.
Claude also bills cache writes: 1.25× the input price for a 5-minute cache, 2× for one hour.
Checked on October 8, 2026 against the pricing pages of Anthropic.
Price History
No price changes since launch
Strengths
- Anthropic's current fast tier, still on the previous tokenizer
- Cached input at $0.10/M
- 50% Batch API discount ($0.50 / $2.50)
Avoid For
- Contexts above 200K tokens
- Pure cost minimisation — GPT-6 Luna and Gemini 3.1 Flash-Lite list far lower