Claude Haiku 5.5
claude-haiku-5-5Run Claude Haiku 5.5 in a hosted Python Task with structured output, web search, extended thinking, file inputs and MCP connectors.
Anthropic's fastest model for high-volume work such as classification, extraction, and routing. It has a 1M token context window, supports adaptive thinking with effort control, and costs $0.10/$0.50 per million tokens for prompts up to 100K tokens. Longer prompts use a higher rate.
What Claude Haiku 5.5 supports in qbash
- Structured outputReplies as JSON that matches a schema you pass.
- MCP connectorsCalls tools on MCP connectors during a request.
- Web searchSearches the web while it answers.
- Extended thinkingReasons through a problem before it replies.
- Reasoning effort5 levels, from low to max.
- Image inputReads WEBP, PNG, JPEG and GIF images.
- Document inputReads PDF, TEXT, DOC, XLS, RTF and 7 more files.
- Prompt cachingRepeated input is billed at the cached rate.
What is qbash?
qbash is where you build Tasks: Python programs that run a business process on managed infrastructure. Your code runs the steps and calls a model like Claude Haiku 5.5 only at the ones that need judgment, with integrations, memory and files available in the same script.
A developer writes the Task once, and anyone on the team runs it from a form, on a schedule, from a webhook or from chat. Each Task is versioned and every run is logged.
Specifications
- Context window
- 1Mtokens
- Max output
- 128,000tokens
Structured output with Claude Haiku 5.5 in a Python Task
qbash.ai.raw calls Claude Haiku 5.5 from a Task. Pass a JSON Schema as output_schema and the reply comes back as a Python dict that matches it, ready to route to a connector.
ticket = qbash.inputs.get("ticket")
triage = qbash.ai.raw(
f"Summarize this support ticket and rate its urgency:\n\n{ticket}",
model="claude-haiku-5-5",
output_schema={
"type": "object",
"properties": {
"summary": {"type": "string"},
"urgency": {"type": "string", "enum": ["low", "medium", "high"]},
},
"required": ["summary", "urgency"],
},
)
# triage is a dict, e.g. {"summary": "...", "urgency": "high"}
if triage["urgency"] == "high":
qbash.integrations.slack.create_message(
channel="#support",
text=triage["summary"],
)
qbash.output(triage)
Connect Claude Haiku 5.5 to MCP servers
In qbash chat and a Task's inline prompts, you can attach MCP connectors to Claude Haiku 5.5. It calls their tools during the request, for example to read a Postgres table or post to a Slack channel.
This is MCP on the model side. Adding MCP servers to Claude Code or Claude Desktop is a separate setup, and qbash can publish your Tasks as an MCP server for those clients as well.
Use Claude Haiku 5.5 with web search
Turn on Web Search in an inline prompt's model bar and Claude Haiku 5.5 searches the web while it answers. Searches are billed at $10.00 per 1,000.
Reasoning effort with Claude Haiku 5.5
Set Effort to low, medium, high, xhigh and max in the prompt's Settings. The default is medium. Higher effort spends more tokens on reasoning before the reply.
Turn on Thinking in the same Settings popover to have Claude Haiku 5.5 reason through a problem before it replies.
Claude Haiku 5.5 pricing
| Input | $0.10 per 1M tokens |
|---|---|
| Output | $0.50 per 1M tokens |
| Cache read | $0.01 per 1M tokens |
| Cache write, 5 minutes | $0.125 per 1M tokens |
| Cache write, 1 hour | $0.20 per 1M tokens |
| Web search | $10.00 per 1K searches |
| Input, above 100K input tokens | $0.50 per 1M tokens |
| Output, above 100K input tokens | $2.50 per 1M tokens |
Standard rates in USD.
Supported file formats
Images
- WEBP up to 5 MB
- PNG up to 5 MB
- JPEG up to 5 MB
- GIF up to 5 MB
Documents
- TEXT
- DOC
- XLS
- RTF
- DOCX
- XLSX
- PPTX
- PPT
- ODT
- ODP
- ODS