Claude Haiku 5.5 for Video Scripts: API Pricing and a Real Pipeline Test
Claude Haiku 5.5 is Anthropic's small, fast tier, and on our script probe it wrote a 30-second Short's script for about a hundredth of a cent in 2.1 seconds. Here is the rate card, the same prompt on Claude Sonnet 5 for comparison, the fact both models got wrong, and how the model fits a pipeline that writes, narrates, renders, and publishes.
Claude Haiku 5.5 is Anthropic's small, fast tier of the Claude 5 family. It reached OpenRouter on 7 October 2026 and joined the model picker on Treza's Text Generation node on 27 October, at a listed rate one tenth of what Claude Haiku 4.5 charges. In a video pipeline, a language model's job is usually the script: the hook, the beats, the on-screen captions, and the call to action that every later node narrates, renders, and cuts. That is the job we tested it on.
This post covers what Claude Haiku 5.5 costs per million tokens and per script, how it compared with Claude Sonnet 5 on an identical prompt, where its script went wrong, and how to run it as the writing step of a pipeline that narrates, renders, captions, and publishes. Every price is the listed rate as of October 2026, and every number from the test is what the provider billed and what the clock said.
The rate card
| Claude Haiku 5.5 | Claude Sonnet 5 | Claude Haiku 4.5 | |
|---|---|---|---|
| Input, per million tokens | $0.10 | $2.00 | $1.00 |
| Output, per million tokens | $0.50 | $10.00 | $5.00 |
| Context window | 1,000,000 tokens | 200,000 tokens | 200,000 tokens |
| Image input | Yes | Yes | Yes |
One caveat. Past 100,000 prompt tokens in a single request, Haiku 5.5 bills at $0.50 input and $2.50 output per million, five times the base rate. A script prompt is a few hundred tokens and never gets near that line; a request that pastes in several hours of transcript could cross it.
The test: one prompt, two models
Every language model that joins the Text Generation node gets the same probe before it is listed: the node's own request with the node's defaults (temperature 0.7, a 1,024 token cap, reasoning off), the system prompt "You write short-form video scripts", and this user prompt:
Write a 30-second vertical video script about why octopuses have three hearts. Use exactly these sections: HOOK (one line), BEATS (4 numbered lines, each under 15 words), ON-SCREEN TEXT (4 captions, each under 6 words), CTA (one line). No emojis, no markdown headers.
We ran it on Claude Haiku 5.5 and on Claude Sonnet 5, the node's default model, on the same morning.
| Claude Haiku 5.5 | Claude Sonnet 5 | |
|---|---|---|
| Prompt tokens | 123 | 122 |
| Reply tokens | 216 | 218 |
| Billed by the provider | $0.00012 | $0.0024 |
| Wall time | 2.1 s | 3.5 s |
| Finish | Stopped on its own, all four sections present | Stopped on its own, all four sections present |
Twenty times cheaper and a second and a half faster, with the same structure held. Neither model spent reasoning tokens or hit the token cap.
What Haiku 5.5 wrote
The reply, unedited:
HOOK Octopuses have three hearts, and two of them stop beating when they swim.
BEATS
- Two hearts pump blood through the gills, one per side.
- The third heart circulates blood to the rest of the body.
- Their blood uses copper, not iron, so it's bluish.
- Swimming starves the body heart, so octopuses prefer crawling.
ON-SCREEN TEXT Three hearts, one body Two hearts stop when swimming Copper blood turns octopus blue Crawling beats swimming for stamina
CTA Follow for more weird animal biology, and tell me which creature to cover next.
The format held completely. Four beats, the longest ten words against a limit of fifteen. Four captions, the longest five words against a limit of six. No emojis, no headers, and the section labels exactly as asked, which lets a downstream node split the reply without a parser. About 65 words of narration, a comfortable 30 second read.
The fact it got wrong
The hook says two of the octopus's three hearts stop when it swims. That is backwards. The standard account is that the two branchial hearts keep pushing blood through the gills, and the single systemic heart, the one that serves the body, is what stops during swimming, which is why octopuses prefer to crawl. Haiku's own fourth beat has it right. Its hook and its second caption contradict its own script.
Sonnet 5 made the same mistake in its hook and repeated it in a beat, so this is not a small-model problem. It is an unchecked-fact problem, and it would have gone straight to narration on either model.
The pipeline fix is cheap, and Haiku 5.5's price makes it cheaper. Wire a second Text Generation node after the writer with a system prompt that checks the hook against the beats and answers with a corrected hook, then feed that into the text to speech step. On Haiku 5.5 that check costs about what the script did. The faceless script guide covers the system prompt rules that keep a channel's scripts on the rails in the first place.
The same model on an image
The Text Generation node takes a Media input, so the second probe wired in a product photo of a folded navy linen shirt on an oak table and asked for a description. Haiku 5.5 came back in 5.9 seconds for $0.00028, almost all of it the 2,147 prompt tokens the image occupied, and the description was accurate: collar, chest pocket, side lighting, shallow depth of field, the weave and the wood grain. That is what the Media input exists for: describe an image, check a generated clip against its prompt, or score a batch of frames against the brief, at a fraction of a cent per clip.
Where the script cost sits in a Short
A script on Claude Haiku 5.5 cost $0.00012. On Claude Sonnet 5 it cost $0.0024. The Story Video starter in chat is priced at $0.78, and the median charged cost of one video generation on Treza is $1.06. Against either number, both script prices are rounding error. A channel writing one script a day for a year spends about $0.04 on Haiku 5.5 and about $0.88 on Sonnet 5.
So the per-script price does not pick the writing model. Latency does: a script that arrives in two seconds instead of three and a half shortens every run. And call count does. A finished pipeline usually runs several text nodes, the writer, a node that turns the transcript into a title, description and tags, a picker that answers in JSON with clip ranges, a judge on the output. Run those on a schedule, or fan a batch out over a wired list with the Run Pipeline node, and that is where a twenty-fold rate difference becomes visible.
On this probe we found no evidence that the dearer model wrote the better script. Sonnet 5 stays the node's default because it carries the production record: in September 2026, 1,100 of the 1,173 Text Generation nodes that ran on Treza used it. Haiku 5.5 is the model to try when a text node runs often, answers in a fixed shape, or has to be fast.
Running Haiku 5.5 as the script step
On a Treza AI video pipeline, the script is the first node in the chain. Set the Text Generation node's model to Claude Haiku 5.5, put the channel's angle and voice in its system prompt, and turn on Avoid recent repeats so a scheduled channel does not rewrite the video it published last Tuesday; the node remembers up to 50 previous runs. Set a fallback model so a call that fails after retries moves to another model rather than failing the run.
From there the script feeds the rest of the line. A text to speech node reads it in one of 82 built-in voices across four models, with 14 languages on the multilingual model, the setup the AI voiceover generator page walks through. Each scene renders as its own clip on the video model you pick, the Sequence node stitches up to 12 shots against the narration, Whisper transcribes the result with word-level timestamps, captions are burned in, and a YouTube or TikTok node publishes it with the title, description and tags written from the transcript. The faceless video generator page covers that build, and the script to video page covers the case where the script already exists and the model's job is to split it into scenes.
Publish the pipeline and it becomes a versioned endpoint. Your application calls a typed /invoke route with JSON in and JSON out, or points an existing OpenAI SDK at the OpenAI-compatible route, and our hosted MCP server exposes the same pipeline as a tool an agent in Claude or ChatGPT can call. When the next small model undercuts Haiku 5.5, you change the dropdown and republish, and the endpoint your product calls does not change. The node's language models sit alongside the video and image models in the model library.
What you actually pay on Treza
A Text Generation node through a Treza pipeline is metered from a prepaid credit balance. One credit equals $0.01, credit packs start at $5, credits never expire, and a failed generation charges nothing. Before a run starts, the platform estimates the cost of the graph and refuses to start one your balance cannot cover. A text node is quoted at a token budget drawn from production history; at Haiku 5.5's rate that hold is under a cent, and the node settles at what the provider billed, with the difference returned to the balance.
For a finished video costed end to end, see the AI video cost breakdown. To try the model on your own brief, open the AI video generator, set the script node to Claude Haiku 5.5, and run one.
Frequently Asked Questions
How much does Claude Haiku 5.5 cost per million tokens?
$0.10 per million input tokens and $0.50 per million output tokens on the endpoint we run, as of October 2026. Past 100,000 prompt tokens in one request the rate is $0.50 and $2.50. That is one tenth of Claude Haiku 4.5's listed rate and one twentieth of Claude Sonnet 5's.
How much does a video script cost on Claude Haiku 5.5?
Our 30-second Short script, a 123 token prompt into a 216 token reply, was billed $0.00012 by the provider and took 2.1 seconds. The same prompt on Claude Sonnet 5 was billed $0.0024 and took 3.5 seconds. On Treza the charge comes from a prepaid credit balance, and a failed call charges nothing.
Is Claude Haiku 5.5 good enough to write video scripts?
On our probe it held every structural rule in the prompt: four beats under fifteen words, four captions under six, labelled sections, no emojis, a 30 second read. It also put a wrong fact in the hook that its own fourth beat contradicted. Claude Sonnet 5 made the same mistake on the same prompt, so the lesson is to put a fact check in the graph rather than to pay more for the writer.
How does Claude Haiku 5.5 compare with Claude Sonnet 5 for scripts?
Twenty times cheaper per script and about forty percent faster on our test, with the same four sections and the same factual error. Sonnet 5 remains the Text Generation node's default because it carries the production record. Haiku 5.5 suits nodes that run often, answer in a fixed shape, or have to be fast.
Can Claude Haiku 5.5 read images?
Yes. Wired into the Text Generation node's Media input, it described a product photo accurately in 5.9 seconds for $0.00028, with the image accounting for about 2,100 of the prompt tokens. That makes it a cheap judge for checking a generated clip against its prompt.
How do I use Claude Haiku 5.5 in a video pipeline through an API?
On a Treza pipeline, Claude Haiku 5.5 is a choice in the Text Generation node's model picker. Wire that node to text to speech, video generation, captions and publishing, publish the pipeline, and call it as a versioned HTTP endpoint with a typed /invoke route and an OpenAI-compatible route. A hosted MCP server exposes the same pipeline as a tool for agents. Generations are metered from prepaid credits that start at $5 and never expire.
