Test fixture
Throughput and TTFT-focused generation tasks.
The model receives the prompt (and optional system message). The run uses scorer rubric_json_metrics with the JSON configuration below. Pass/fail and partial credit are determined entirely by that scorer against the model output; no human grading.
Return JSON only with keys benchmark, title, notes, done. Set benchmark to "speed", title to "Code Review Notes", and done to true. Generate 24 concise code-review notes for performance, reliability, observability, and maintainability.
{
"metrics": {
"complete": {
"checks": [
{
"contains": [
"speed"
]
},
{
"contains": [
"done"
]
},
{
"contains": [
"code"
]
}
]
}
}
}temperature
0
max_tokens
900
timeout (s)
180
type
throughput
file
speed-code-review-notes.json