Claude Haiku 4.5 · Anthropic

What does Claude Haiku 4.5 cost to run in a coding agent?

Claude Haiku 4.5 costs about $0.022 per average coding-agent request on our measured traffic, at list prices of $1 per million input and $5 per million output tokens. It scores 17 on the Artificial Analysis Intelligence Index, 42nd of the 48 models we track. Source: Artificial Analysis (artificialanalysis.ai)

List prices: Anthropic pricing, checked 2026-10-13. Request costs: Standard Compute gateway data, 79,301 coding-agent requests from 75 users, Sep 13 to Oct 12, 2026. How we compute this

A newer version is out. Claude Haiku 5.5 →

Run Claude Haiku 5.5 in your agent

Or let the router send each step to the cheapest model that can do it, on providers we vet for price, uptime and speed. Plans from $19 a month.

See plans

What Haiku 4.5 costs per coding-agent request

Cost at list price	Haiku 4.5
Prompt under 100K tokens50% of requests, 87% of input cached	$0.016
Prompt of 100K to 272K39% of requests, 88% of input cached	$0.033
Prompt over 272K tokens11% of requests, 86% of input cached	n/a
Average request	$0.022
100 requests	$2.22
Median session16 requests; a session ends after 10 idle minutes	$0.27
A month of regular usemedian of 20 users active on 5+ days	$24.74

Each logged request priced at the vendor’s list rates for its own token counts, including cache reads and any higher rate for long prompts. Average output: 698 tokens per request. 21% of requests were larger than Haiku 4.5's 200K context window and are left out of its figures.

Haiku 4.5 benchmarks, speed and latency

Artificial Analysis	Value	Rank
Intelligence Index10 evaluations, 0 to 100	17	42 of 48
Terminal-Bench 4.0agentic coding in a terminal	0.0%	35 of 39
Output speedmedian	95 tokens/s	22 of 44
Time to first answer tokenincludes thinking time	22 s	10 of 44

Source: Artificial Analysis (artificialanalysis.ai). Measured variant: Claude 4.5 Haiku (Reasoning), speed on Anthropic. Fetched October 13, 2026. Rank is among the models on this site, older versions included. Artificial Analysis runs Terminal-Bench 4.0 in its own setup, so its scores differ from the tbench.ai leaderboard.

What to watch for

Claude Haiku 5.5 has replaced it and costs $0.013 per request on our traffic, 41% less. Its 200K-token window is smaller than 21% of our coding-agent requests, which are left out of the figures above, so long agent sessions will hit the limit. It is weak on long agentic coding tasks (see Terminal-Bench 4.0 above), so a cheap request can turn into more steps and retries.

About Claude Haiku 4.5

Claude Haiku 4.5 is Anthropic's previous small model, released October 15, 2025 for cheap, fast calls such as file triage, summarization, routing and subagents. Haiku 5.5 replaced it in October 2026. It scores 17 on the Artificial Analysis Intelligence Index and 0% on Terminal-Bench 4.0, both in Artificial Analysis's run and by Anthropic's count. Its 200K context is smaller than 20.7% of the coding-agent requests we measured, and it lists at $1/$5 per million tokens, ten times Haiku 5.5's base rate.

Sources: Artificial Analysis

Claude Haiku 4.5 FAQ

How much does Claude Haiku 4.5 cost per coding-agent request?

$0.022 on average at list price, $2.22 per 100 requests, and $0.27 for a median 16-request session. That is 79,301 real coding-agent requests (Sep 13 to Oct 12, 2026) priced at Anthropic's list rates; the median request sent 99,534 input tokens, 87% of input was read from cache, and requests returned 698 output tokens on average.

Does Claude Haiku 4.5 charge more for long prompts?

No. Claude Haiku 4.5 bills the same rate per token up to its 200K-token context window, which is smaller than 21% of the coding-agent requests we measured.

How fast is Claude Haiku 4.5?

Artificial Analysis measures Claude 4.5 Haiku (Reasoning) at 95 output tokens per second, with 22 seconds to the first answer token, thinking time included, on Anthropic's API. Source: Artificial Analysis (artificialanalysis.ai).

Is Claude Haiku 4.5 good for AI agents?

Short, simple subtasks in setups that still run Haiku 4.5. On our scoring, which is calibrated to public benchmarks, it rates 4.4/10 for agentic ability and 4.5/10 for output quality. Flat pricing with no long-prompt surcharge.

What is Claude Haiku 4.5's context window?

Claude Haiku 4.5 has a 200K-token context window. Anthropic is the maker.

Can I self-host Claude Haiku 4.5?

No. Claude Haiku 4.5 is a proprietary, hosted-only model from Anthropic. You access it through an API, such as OpenRouter, rather than running the weights yourself.

What are the downsides of Claude Haiku 4.5?

The main trade-offs: Superseded by Claude Haiku 5.5, which is cheaper below 100K tokens and scores far higher; 200K context, smaller than 20.7% of the coding-agent requests we measured; 0% on Terminal-Bench 4.0, per Anthropic. It's strongest for short, simple subtasks in setups that still run Haiku 4.5.