AI Index

Claude Haiku 4.5

Language Anthropic Verified Released Oct 15, 2025
  • In TextImage Out Text
  • Reasoning
  • Tool use
  • Structured output

Prices

5 offerings · prices as listed
Claude API Official Global · USD claude-haiku-4-5
Conditions/ 1M tokensInputCache readCache writeOutput
5-minute cache1-hour cache
Standard $1.00$0.10$1.25$2.00$5.00
Batch $0.50———$2.50

Prices from Claude API pricing (checked Nov 8, 2026)

Standard: the price a request with no conditions pays (the headline price elsewhere on the site).

Details

Knowledge cutoff
Feb 2025
Open weights
—
Family
claude-haiku
Model ID

Not stated on Anthropic’s pages: max input, reasoning efforts, reasoning can be off, max thinking budget, built-in tools, aliases.

Cited scores

From third-party leaderboards, not measured by us
  • Epoch Capabilities Index 142.8

    #89 of 118 listed models · CI 141.1–144.4

  • Epoch AI · Chess Puzzles 8%

    #81 of 85 listed models · 59K thinking budget · ±2.73

    Also listed: 32K thinking budget 5%

  • Epoch AI · GPQA Diamond 71.2%

    #94 of 168 listed models · 32K thinking budget · ±3.23

    Also listed: claude-haiku-4-5-20251001 60.5%

  • Epoch AI · MATH Level 5 96.4%

    #9 of 64 listed models · 32K thinking budget · ±0.52

    Also listed: claude-haiku-4-5-20251001 86.9%

  • Epoch AI, 'Capabilities & benchmarking'. Published online at epoch.ai. Retrieved from 'https://epoch.ai/benchmarks'. Licensed CC BY 4.0. · License · as of Oct 18 – Nov 5, 2026
  • EQ-Bench by Samuel J. Paech (eqbench.com; EQ-bench/EQ-bench-site on GitHub), MIT License. · License · as of Oct 24 – 28, 2026
  • changes we made

Provider-reported scores

Published by Anthropic, under the conditions it states
BenchmarkSettingScore
SWE-bench Verifiedbash + string-replace edit tools, averaged over 50 trials, 128K thinking budget, prompt addition encouraging tool use73.3%
τ²-benchRetail domain; extended thinking (128K budget), averaged over 10 runs, prompt addendum83.2%
Airline domain; extended thinking (128K budget), averaged over 10 runs, prompt addendum63.6%
Telecom domain; extended thinking (128K budget), averaged over 10 runs, prompt addendum83%
OSWorld-VerifiedOSWorld-Verified framework, 100 max steps, averaged over 4 runs, 128K total / 2K per-step thinking budget50.7%
AIME 2025with Python; pass@1 over 16 trials averaged over 10 runs, 128K thinking budget96.3%
no tools; pass@1 over 16 trials averaged over 10 runs, 128K thinking budget80.7%
GPQA Diamondaveraged over 10 runs, 128K thinking budget73%
MMMLUaverage of 10 runs over 14 non-English languages, 128K thinking budget83%
MMMUvalidation set, averaged over 10 runs, 128K thinking budget73.2%

Included in coding plans

  • GitHub Copilot Pro $10.00 / month
    1,500 credits per month
    Variable additional usage on top of your base. Flex allotments may change over time.
    15 USD of usage per month
  • GitHub Copilot Pro+ $39.00 / month
    7,000 credits per month
    Variable additional usage on top of your base. Flex allotments may change over time.
    70 USD of usage per month
  • GitHub Copilot Max $100.00 / month
    20,000 credits per month
    Variable additional usage on top of your base. Flex allotments may change over time.
  • GitHub Copilot Business $19.00 / month
    1,900 credits per month
  • GitHub Copilot Enterprise $39.00 / month
    3,900 credits per month
  • Kiro Pro $20.00 / month
    1,000 credits per month
  • Kiro Pro+ $40.00 / month
    2,000 credits per month
  • Kiro Pro Max $100.00 / month
    5,000 credits per month
  • Kiro Power $200.00 / month
    10,000 credits per month

All coding plans →

Changes

All changes →
  1. Now priced on Amazon Bedrock Input $1.00 · Output $5.00 · Input (Batch) $0.50 · Output (Batch) $2.50 · +3 more
  2. 3 new prices on Claude API Cache write (1-hour cache) $2.00 · Input (Batch) $0.50 · Output (Batch) $2.50

Sources

Anthropic’s official policies: data use, privacy, terms →

Anthropic · United States · anthropic.com

Snapshots, the same family and other models from this provider