claude-3-haiku Benchmarks

Provider: anthropic

Explore real-world latency and throughput results for claude-3-haiku. These measurements come from automated benchmarking runs against the provider APIs using the same harness that powers the public cloud dashboard.

Want a broader view of this vendor? Visit the anthropic provider hub to compare every tracked model side-by-side.

Visit anthropic Official Website

Benchmark Overview

Avg Tokens / Second

87.20

Avg Time to First Token (ms)

0.32

Runs Analysed

903

Last Updated

Oct 18, 2025, 12:23 AM

Key Insights

  • claude-3-haiku streams at 87.20 tokens/second on average across the last 903 benchmark runs.

  • Performance fluctuated by 92.00 tokens/second (105.5% coefficient of variation), indicating variable behavior across benchmark runs.

  • Average time to first token is 0.32 ms (excellent latency), suitable for latency-sensitive workloads.

  • Latest measurements completed on Oct 18, 2025, 12:23 AM based on 903 total samples.

Performance Distribution

Distribution of throughput measurements showing performance consistency across benchmark runs.

Performance Over Time

Historical performance trends showing how throughput has changed over the benchmarking period.

claude-3-haiku

Benchmark Samples

ProviderModelAvg Toks/SecMinMaxAvg TTF (ms)
anthropicclaude-3-haiku87.2016.00108.000.32

Frequently Asked Questions

The latest rolling average throughput is 87.20 tokens per second with an average time to first token of 0.32 ms across 903 recent runs.

Benchmarks refresh automatically whenever the monitoring cron runs. The most recent run completed on Oct 18, 2025, 12:23 AM.

Related Links