Playground

200K ctx · $5.00/$15.00 per 1M

System instructions

45 estimated tokens. Request preview only; instructions do not change the canned reply.

Sampling

Temperature and Top p appear in the request preview; they do not change simulated content.

Temperature

0.2

Top p

Response format

Text and JSON use the same sample answer. JSON wraps it in an object.

Tools

Delivery

Stream responseTokens render as they arrive instead of on completion.

Grounded ops

How does Embedding 3 latency compare to the 30-day baseline?

Model 4.5· inspecting

Embedding 3 holds a p50 of 94 ms against a 30-day baseline of 98 ms — a small but real improvement. Median latency: 94 ms (-4 ms vs baseline) 95th percentile: 138 ms, inside the 150 ms SLO Throughput: 1.8k tps sustained, no spikes over 100 ms 2,840 requests in the window, 0 retries No error spikes above the baseline. The service is comfortably inside its SLO.

Completed · 2.14 s · 386 tokens · $0.0122

0 tokens

Enter a prompt to enable Run.