We use cookies to enhance your experience, provide personalised content, and analyse our traffic; see our Cookie Policy.
anthropic/claude-haiku-5-5Near-frontier answers at a nano price
Anthropic's fastest model, released 7 October 2026, and the largest quality jump they have shipped in a point release: 45.9% on Humanity's Last Exam against Haiku 4.5's 10.2%, and 1620 on GDPval-AA against 735. That puts it within reach of Sonnet 5.5 while costing about a tenth of Haiku 4.5 per token. On Tchavi it is 0.24 credits per 1K in and 1.2 out — priced like a small model, answering like a large one. It thinks before it replies by default, so give it a task worth thinking about.
Built for replies a person is waiting for
Reasons before answering, with no flag to set
Send a scan or a screenshot alongside your text
Function calling and guaranteed structured output
Haiku 5.5, in almost every case. It is cheaper per token and scores several times higher on Anthropic’s own benchmarks. Stay on 4.5 only if you depend on setting temperature.
The model holds a million tokens and Tchavi serves all of them. Anthropic charges five times more for any prompt over 100,000, and we pass that tier through instead of capping the window or raising the price for everyone: below the line you pay 0.24 credits per 1K in and 1.2 out, above it 1.2 and 6. Cached tokens count toward the threshold.
It reasons before replying and that thinking is billed as output, even though you do not see it. It is usually still cheaper than sending the task to a bigger model.
Not through Tchavi today. Anthropic allows it only at some effort levels, and the model can skip a needed tool call when thinking is disabled, so we leave it on.
Yes. Top up your Tchavi account with Orange Money, MTN MoMo, M-Pesa or another mobile money operator. No bank card is needed. Credits land as soon as the payment is confirmed and are deducted as you use them.
Yes. Keep your current OpenAI SDK, just change the base URL to Tchavi's (shown in the API tab) and use your Tchavi key. The rest of your code stays the same.
import Tchavi from '@tchavi/sdk';
const client = new Tchavi({ apiKey: 'API_KEY' });
const response = await client.chat.completions.create({
model: 'claude-haiku-5-5',
messages: [
{
"role": "user",
"content": "Hello!"
}
],
});
console.log(response.choices[0].message.content);
console.log('Credits used:', response.tchavi.credits_used);Every parameter this model accepts on the API. Check the SDK for full typings.
modelRequiredID of the chat model (e.g. `claude-haiku-5-5`).
messagesRequiredConversation messages. Each item has a `role` (system/user/assistant) and a `content` that's either a plain string or an array of content blocks (`text`, `image_url`, `document_url`).
temperature1Sampling temperature between 0 and 2. Higher values = more random.
max_tokensMaximum number of tokens to generate in the response.
streamfalseIf true, partial message deltas are streamed as Server-Sent Events.
toolsFunction-calling toolbox. Pass an array of `{ type: "function", function: { name, description, parameters } }`. Returned tool calls land in `choices[].message.tool_calls`. Only models with `supportsTools: true` accept this.
tool_choiceHow the model picks among tools: `"auto"` (default), `"none"` to force a plain text reply, or `{ type: "function", function: { name } }` to force a specific tool. Only with `supportsTools: true`.
response_formatConstrain the output. Pass `{ type: "json_object" }` to force valid JSON (the model still needs a JSON instruction in your prompt). Only models with `supportsJsonMode: true` honour this.
top_p1Nucleus sampling: only tokens in the top `top_p` probability mass are considered.
stopUp to 4 sequences where the API will stop generating further tokens.
frequency_penalty0Between -2.0 and 2.0. Positive values penalize tokens that have appeared frequently.
presence_penalty0Between -2.0 and 2.0. Positive values encourage the model to talk about new topics.
messages[].content[].image_urlAttach an image on a user message. Pass a content block `{ "type": "image_url", "image_url": { "url": ... } }`. PNG / JPEG / WebP, up to 4 per request. Only models with `supportsImageInput: true` accept this block (e.g. `gpt-4o`, `claude-opus-4-7`).
messages[].content[].document_urlAttach a PDF on a user message. Pass a content block `{ "type": "document_url", "document_url": { "url": ... } }`. Up to 5 PDFs per request. Only models with `supportsDocuments: true` accept this block (e.g. `claude-sonnet-4-6`, `gpt-4.1`).
effort"high"Reasoning depth on Anthropic adaptive-thinking models (Opus 4.7+ / Sonnet 4.6+). Values `"low"`, `"medium"`, `"high"`. Higher = more deliberate, slower. Defaults to `"high"` on Opus 4.8.
This model costs more upstream once a prompt passes a threshold, and we pass that through rather than cap the window or raise the price for everyone. The rate above is the first band.
| Prompt length | Input / 1K | Output / 1K |
|---|---|---|
| Under 100K tokens | ||
| 100K tokens and above |
The threshold counts every input token on the request, cached ones included — a 95K cache hit plus 10K of new prompt is a 105K prompt. Output is priced by the prompt’s length, so a long prompt with a short answer still pays the higher output rate.
Estimated over 30 days, for a typical message: ~500 input + ~150 output tokens, billed to the thousandth of a credit.