NC Bench
Overview
About
Tests
Models
Vendors
Run: 2026-02-20-01KHY8KEW50EZ40C5WS6A5G18K
Run Details
Model:
Gemma 3 4B
Test:
N-Length Sentences
Scenario:
Write sentences with 10 words each
Duration:
1.6s
Cost:
$0.000012
Provider:
DeepInfra
Evaluation summary
Score
Evaluator
Details Link
79.84%
Matches word count
(details)
Model input
User/Human
Please write the opening scene of a murder mystery where each sentence has exactly 10 words. Please make sure that each sentence has the same length. Only return the opening itself.
Model response
Messages
Meta
AI
Rain lashed against the windows, a dismal, gray hue. Detective Harding surveyed the opulent, unsettling study now anew. Crimson stained the Persian rug, a terrible, shocking sight. A priceless antique clock stood frozen, silent, and quite still. The victim lay motionless, a wealthy man, clearly distressed. His face was pale, reflecting the room’s dim, mournful light. A single gunshot echoed, a devastating, terrible sound now. Harding touched the cold hand, confirming life’s final plea. “This murder scene demands careful study, I must confess.”
created_at
"2026-02-20T19:32:07.185Z"
2/20/2026, 7:32:07 PM
model
"google/gemma-3-4b-it"
app_id
182717
external_user
(null)
streamed
true
cancelled
false
latency
163
moderation_latency
(null)
generation_time
1402
tokens_prompt
63
tokens_completion
142
native_tokens_prompt
46
native_tokens_completion
121
native_tokens_completion_images
(null)
native_tokens_reasoning
0
native_tokens_cached
0
num_media_prompt
(null)
num_input_audio_prompt
(null)
num_media_completion
0
num_search_results
(null)
origin
"https://ncbench.com/"
is_byok
false
finish_reason
"stop"
native_finish_reason
"stop"
usage
0.00001152
router
(null)
provider_responses
0
endpoint_id
"d3de3bd4-81bc-48fb-924f-2a87b2a36e75"
id
"chatcmpl-Rxa40kH2xgfGpZQGeqyrjNoB"
is_byok
false
latency
59
model_permaslug
"google/gemma-3-4b-it"
provider_name
"DeepInfra"
status
200
api_type
"completions"
id
"gen-1771615927-2oj57CdHFX5BcluGiHFn"
upstream_id
"chatcmpl-Rxa40kH2xgfGpZQGeqyrjNoB"
total_cost
0.00001152
cache_discount
(null)
upstream_inference_cost
0
provider_name
"DeepInfra"
Evaluation details
Result
Evaluator
Details
Meta Data
79.84%
Matches word count
n/a
neededClean
false
sentences
9
wordCounts
0
9
1
9
2
9
3
10
4
9
5
10
6
9
7
9
8
9