Claude Haiku 4.5 — speed/cost analysis

End-of-lesson check

15 questions · take it digitally for instant feedback at tendril.neural-forge.io/learn/quiz/end-modelx-claude-haiku-45-builders

1. A developer is building a ticket routing system that must handle 10,000 requests per minute. Which model would be most cost-effective for the initial classification step?

   1. Claude Haiku 4.5, because its low cost per million tokens makes it economical for high-volume classification tasks
   2. Claude Opus, because it provides the best latency for real-time applications
   3. Claude Sonnet 4.6, because its higher quality ensures accurate routing decisions
   4. Either model would work equally well since both can handle this QPS

2. What does the lesson mean when it says Haiku is 'the quiet workhorse'?

   1. Haiku has the lowest market share of any Anthropic model
   2. Haiku requires the least amount of computational resources to run
   3. Haiku operates silently without generating any log output
   4. Haiku is underappreciated but handles high-volume workloads reliably and affordably

3. A student asks what 'latency' means in the context of AI models. Which definition is correct?

   1. The total number of tokens processed in a request
   2. The time between sending a request and receiving the first response
   3. The cost charged by the API for processing a request
   4. The maximum throughput the model can handle per second

4. What does QPS stand for in the lesson's comparison table?

   1. Question Processing System
   2. Queries Per Second
   3. Quality Performance Score
   4. Quantum Processing Speed

5. A developer implements a system where Haiku handles initial document parsing, and if confidence is low, it escalates to Sonnet. What is this architectural pattern called?

   1. Cascading
   2. Checkpointing
   3. Load balancing
   4. Failover

6. A product manager wants to reduce API costs by 80% while maintaining quality on complex queries. Which approach does the lesson recommend?

   1. Use Haiku for all requests and accept lower quality
   2. Switch to a different AI provider entirely
   3. Route simple requests to Haiku and escalate complex ones to Sonnet
   4. Use Sonnet for all requests since it has better quality

7. Which task is the lesson LEAST likely to recommend Haiku for?

   1. Classifying support tickets by department
   2. Autocomplete suggestions in a search bar
   3. Writing a long-form analytical essay
   4. Extracting structured data from invoices

8. A company processes 1 million customer messages per day. Why might Haiku help their bottom line more than Sonnet?

   1. Haiku provides better customer support features
   2. Haiku is cheaper per token and optimized for high-volume workloads
   3. Haiku has higher quality so customers are more satisfied
   4. Haiku requires fewer API calls per message

9. The lesson mentions 'structured extraction from semi-clean docs.' What type of document would be most suitable for Haiku?

   1. A scanned PDF with complex tables and merged cells
   2. A handwritten letter with scribbles and cross-outs
   3. A heavily redacted legal document with blacked-out sections
   4. A completed web form with consistent fields

10. What is the primary reason to use Haiku for 'tool-call decisions in multi-step agents'?

    1. Haiku has special tool-calling features other models lack
    2. Agents need to make many quick decisions and Haiku's low latency keeps the agent responsive
    3. Tool calls require high-quality reasoning only Sonnet provides
    4. Tool calls are expensive and should be minimized

11. Based on the lesson, if a developer's priority is minimum time-to-first-token, which model should they choose?

    1. Claude Sonnet 4.6
    2. Claude Opus
    3. Claude Haiku 4.5
    4. Any Claude model will have similar latency

12. What does the lesson imply about using Haiku for 'autocomplete-style suggestions'?

    1. It requires additional human review for every suggestion
    2. It should be avoided because quality is too low
    3. It is more expensive than other use cases
    4. It works well due to low latency and can return suggestions under 200ms

13. A startup is building their first AI product with a limited budget. Why might the lesson suggest starting with Haiku?

    1. Sonnet is too expensive for any startup to use
    2. Haiku has better documentation for beginners
    3. Haiku is the only model that works for startups
    4. Starting with Haiku reduces costs during development and lets them scale affordably

14. The lesson states Haiku can handle 80% of work while cutting costs. What happens to the remaining 20% of requests?

    1. They are queued for batch processing
    2. They are rejected
    3. They are handled by Sonnet after escalation
    4. They are processed by a different AI provider

15. Why might a company choose NOT to use Haiku for long document summarization?

    1. Haiku is too expensive for summarization
    2. Haiku has lower token limits for input
    3. Haiku does not support summarization
    4. Long documents require reasoning and quality chat, which Sonnet handles better