Claude 3.5 Haiku

Cost & access

Claude 3.5 Haiku is proprietary via Anthropic. Typical billing: input + output tokens; ChatGPT-style subscriptions are separate from API access. With a 200K tokens context window, long PDFs or chat histories increase input tokens quickly—trim history or summarize older turns in production.

Functional understanding

• Low-latency Claude for real-time assistants.
• Modalities: text · License: Proprietary · Released 2024-10-22.
• Best-fit workflows for this model:
• • Drafting, summarization, and structured extraction from long documents.

Technical foundation

• Anthropic reports Undisclosed parameters; training data: Undisclosed.
• Context: 200K tokens. Open weights: no.
• Claude 3.5 Haiku is positioned as a general-purpose model in the Anthropic lineup.

First API call

Install anthropic, set ANTHROPIC_API_KEY, and call Claude 3.5 Haiku with the Messages API.

from anthropic import Anthropic

client = Anthropic()
msg = client.messages.create(
    model="claude-3-5-haiku-20241022",
    max_tokens=1024,
    messages=[{"role": "user", "content": "Summarize MoE vs dense models."}],
)
print(msg.content[0].text)

Important technical topics

• Prompting Claude 3.5 Haiku: be explicit about output format. Weak: "Analyze this." Better: "Return JSON with fields id, total, date for Anthropic billing data."
• Temperature: use 0–0.3 for extraction and compliance on Claude 3.5 Haiku; 0.7–1.0 for brainstorming.
• Tokens: Claude 3.5 Haiku bills by tokens (~¾ word each). Undisclosed parameters affect capability; your bill is driven by context length and call volume.
• Context window (200K tokens): everything in one request—system prompt, tools, RAG chunks, and history—must fit. Truncate or summarize when approaching the limit for Claude 3.5 Haiku.

Real enterprise patterns

• RAG with Claude 3.5 Haiku: retrieve from your vector DB, cite sources in the prompt.
• Tool calling: define JSON schemas; let Claude 3.5 Haiku request functions, not free-form SQL.
• Eval suite: regression prompts before each model or prompt change.
• Cost routing: default to Claude 3.5 Haiku for hard tasks; smaller sibling model for triage.

Production & security

• Secrets: never commit keys for Claude 3.5 Haiku; use vault + per-environment rotation.
• PII: mask before inference; log redacted prompts only.
• Observability: trace id per request; log model=claude-3-5-haiku, tokens in/out, latency.
• Rate limits: handle Anthropic 429/5xx with exponential backoff and circuit breakers.
• Guardrails: schema-validate JSON; block disallowed topics; cross-check numbers against source docs.

Mini projects with this model

• Support copilot: Claude 3.5 Haiku drafts replies from KB snippets.
• Contract clause extractor with human approval.
• Weekly metrics narrative from SQL + CSV exports.
• Agent that files expenses from receipt photos (if multimodal).

Suggested stack

• Language: Python 3.11+
• LLM: Claude 3.5 Haiku through anthropic SDK
• Tool use: Anthropic tool schemas + your FastAPI backends
• UI: Streamlit or Next.js for internal tools
• APIs: FastAPI
• Vector DB (RAG): Pinecone / Chroma / pgvector

Learning path

• Python basics
• HTTP/REST and environment variables
• Anthropic authentication and Claude 3.5 Haiku model id (claude-3-5-haiku-20241022)
• First successful call to Claude 3.5 Haiku
• Prompt design and JSON / structured outputs
• RAG
• Tool use / function calling
• Evals and regression sets
• Production deploy + monitoring