Claude Haiku 4.5
Anthropic · Released Oct 2025
Anthropic's value baseline at $1/$5 per 1M tokens — the fast, low-cost tier rivals benchmark against, with 200K context and 64K output. Best for high-volume latency-sensitive work: classification, extraction, summarization, routing. Reach for Opus 4.8 or Sonnet 5 when deep multi-step reasoning is needed.
Is it right for you?
Good for
- High-volume, latency-sensitive workloads — classification, extraction, routing, summarization, and cheap agent sub-steps at $1/$5 per 1M tokens
- Cost-efficient agent swarms and sub-agents — the budget tier of a Haiku > Sonnet > Opus escalation ladder
- Larger-context budget work — 200K context and up to 64K output, far beyond Haiku 3.5's 8K limit
Not good for
- Complex multi-file architecture or frontier reasoning — route those to Sonnet 5 or Opus 4.8; Haiku is not a frontier model
- Absolute-cheapest coding value in the lightweight tier — Microsoft's MAI-Code-1-Flash claims a 16-point SWE-Bench Pro lead over Haiku 4.5 inside Copilot
How it performs by task
Classification and extraction
Ideal cheap, fast tier for high-throughput structured tasks at $1/$5 per 1M
Summarization and drafting
Fast and affordable for document summarization and content workflows
Coding (lightweight)
Handles routine completions and refactors; MAI-Code-1-Flash claims a 16-point SWE-Bench Pro lead in the lightweight tier
Complex reasoning
Not a frontier model — route deep multi-step reasoning to Sonnet 5 or Opus 4.8
Pricing
Input
$1 / 1M tokens
Output
$5 / 1M tokens
Context
200K tokens
Verdict history
Sources
Verification log
- Pricing— No changes
Automated agent
- Pricing— No changes
Automated agent
- Pricing— Updated
Automated agent
Created new registry entry Claude Haiku 4.5 (was missing): $1/$5 per 1M, 200K context, released 2025-10-15, verdict recommended (budget/value tier referenced by MAI-Code-1-Flash and GPT-5.6 Luna comparisons).
- Profile— No changes
Imported at launch
- Pricing— No changes
Imported at launch