NC Bench
Overview
About
Tests
Models
Model
Claude 3.7 Sonnet
Test
N-Length Sentences
Scenario
Write sentences with 20 words each
Run ID
2025-02-24-01JMX1JC6MQK3FM6FF9C3MS5EX
Model input
User/Human
Please write the opening scene of a murder mystery where each sentence has exactly 20 words. Please make sure that each sentence has the same length. Only return the opening itself.
Model response
Messages
Meta
AI
Detective Sarah Morgan arrived at the crime scene, her breath forming clouds in the cold November air. Yellow police tape fluttered in the breeze, marking off the area where the body had been discovered. The victim lay sprawled on expensive marble tiles, one hand clutching a torn photograph, the other empty. Blood had pooled beneath his head, creating a dark halo that contrasted sharply with his pale skin. Sarah knelt beside the corpse, careful not to disturb potential evidence as she examined the fatal wound. "Blunt force trauma to the back of the head," she muttered, noticing an unusual pattern in the injury. A crystal paperweight lay nearby, its once-pristine surface now smeared with what appeared to be blood. The mansion's opulence felt hollow now, its expensive furnishings mere witnesses to the violent act committed here. "Time of death estimated between midnight and three AM," announced the medical examiner, looking up from below. Sarah noted the victim's watch had stopped at exactly twelve thirty-two, its crystal face shattered beyond repair.
id
"gen-1740436484-KVTmjvlJ3qMDRpywGhIW"
upstream_id
"msg_01S7T27MaRxm2RKwbAxb9rfJ"
total_cost
0.003717
cache_discount
(null)
provider_name
"Anthropic"
created_at
"2025-02-24T22:34:49.419912+00:00"
Monday, February 24th, 2025 at 10:34:49 PM GMT+00:00
model
"anthropic/claude-3-7-sonnet-20250219:beta"
app_id
182717
streamed
true
cancelled
false
latency
353
moderation_latency
(null)
generation_time
3831
tokens_prompt
42
tokens_completion
210
native_tokens_prompt
44
native_tokens_completion
239
native_tokens_reasoning
0
num_media_prompt
(null)
num_media_completion
(null)
num_search_results
(null)
origin
"https://ncbench.com/"
is_byok
false
finish_reason
"stop"
native_finish_reason
"stop"
usage
0.003717
Evaluation details
Result
Evaluator
Details
Meta Data
3.5954%
Matches word count
n/a
neededClean
false
sentences
10
wordCounts
0
17
1
17
2
17
3
17
4
17
5
18
6
17
7
17
8
17
9
18