◈ Multi-Provider · tokens/sec
Models Speed Benchmark
Same prompt · Streaming · Ranked by tokens / second
6 gateways · 30 models · One sentence answer
⏱ Auto-run Schedule
Loading…
Saved to
schedule.json + BENCHMARK_INTERVAL. Use disabled to stop autoruns. Next run shown above when enabled.
⚙ Providers & Models
ID
Provider
Base URL
API Key Env
Tier
★
Actions
Edits save to the server (D1 on Cloudflare,
models.json locally). Drag ⋮⋮ to reorder — top runs first; disabled models are skipped in runs.
Benchmark Output
Running…
—
Models Tested
—
Fastest (tok/s)
—
Winner
Tokens per Second ↑ (higher is better)
Loading results.json …
tokens / second →
Historical Performance — Tokens per Second 3 panels · up to 12 models each · errors = gaps
Run the benchmark multiple times to see historical trends