llm-benchmarks
githubdrose.io

Claude Sonnet 5 Benchmarks


Measured on anthropic

8 runs
anthropic site
Avg Tokens / Second
18.30
Decode Tokens / Second
Avg Time to First Token (ms)
3180.00
Runs Analysed
8
Last Updated
Sep 7, 2026, 08:57 PM

Throughput distribution

Throughput over time

Claude Sonnet 5anthropic27tok/s

Generated throughput, including reasoning tokens — so a thinking model's line does not collapse; the table ranks on visible tokens · one cell per model, showing its best-covered provider · models with at least 4 samples in the window first, fastest of those at the top · shared vertical scale, 0 to 32 tok/s (99th percentile) · dashed rule is the model's own mean · 1 further provider not drawn

Every lane measuring this model

3 measured · live first · a lane is a provider, a catalogue id and a transport
ProviderLaneAvg Toks/SecMinMaxAvg TTF (ms)n
anthropicvia OpenRouter18.309.2428.203180.008
anthropicvia OpenRouter · disabled28.9011.3034.501180.0038
anthropicdirect · disabled36.5036.5036.50620.001

Frequently Asked Questions

How fast is Claude Sonnet 5?
The latest rolling average throughput is 18.30 tokens per second with an average time to first token of 3180.00 ms across 8 recent runs.
How often are these benchmarks updated?
Benchmarks refresh automatically whenever the monitoring cron runs. The most recent run completed on Sep 7, 2026, 08:57 PM.