NC Bench
Overview
About
Tests
Models
Vendors
Run: 2026-02-20-01KHY9SD7R40E9MBY9GM0R1WWT
Run Details
Model:
Gemma 3 4B
Test:
Dialogue tags
Scenario:
Write 200 words with 90% dialogue
Duration:
3.2s
Cost:
$0.000022
Provider:
DeepInfra
Evaluation summary
Score
Evaluator
Details Link
0.00%
Matches word count
(details)
99.49%
Dialogue to Total Word Ratio
(details)
49.7442%
Overall
Model input
User/Human
You are a author of comedic heist books. Please write a 200 word scene that has 90% words of dialogue between two thieves named Molly and Bill who are planning to break into a children's party to steal all the presents. Only include the prose of the scene without any markdown or commentary.
Model response
Messages
Meta
AI
“Right, so, the bouncy castle. Seriously?” Molly asked, tapping a manicured nail against the blueprint. “It’s the distraction, Molly! Kids love bouncy castles. Guaranteed chaos. Think of the cover.” Bill replied, adjusting his ridiculously oversized sunglasses. “Chaos is your specialty, Bill. Mine is precise operation. Do you really think we can just waltz in and grab all the presents?” “Relax. Timed perfectly. Fifteen minutes. That’s all we need. Little Timmy's got that limited edition Lego Millennium Falcon. Worth a fortune.” “A very small, very plastic fortune.” “Details, details! Look, you handle the drone, map the present locations. I’ll manage the… exuberant children.” “’Exuberant children’? You’re terrifying me.” “Don’t be. Just think of it as a very enthusiastic obstacle course.” “What if someone asks where you’ve been?” “I’ll say I was… admiring the balloons. Lots of balloons.” “You’re hopeless.” “Deliciously hopeless," Bill grinned.
created_at
"2026-02-20T19:52:50.706Z"
2/20/2026, 7:52:50 PM
model
"google/gemma-3-4b-it"
app_id
182717
external_user
(null)
streamed
true
cancelled
false
latency
184
moderation_latency
(null)
generation_time
2903
tokens_prompt
93
tokens_completion
285
native_tokens_prompt
75
native_tokens_completion
235
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
0.0000218
router
(null)
provider_responses
0
endpoint_id
"d3de3bd4-81bc-48fb-924f-2a87b2a36e75"
id
"chatcmpl-RgwMRanBld9RCaBdP6HYBw0u"
is_byok
false
latency
68
model_permaslug
"google/gemma-3-4b-it"
provider_name
"DeepInfra"
status
200
api_type
"completions"
id
"gen-1771617170-edj4J0NwExWhl4hfz6d8"
upstream_id
"chatcmpl-RgwMRanBld9RCaBdP6HYBw0u"
total_cost
0.0000218
cache_discount
(null)
upstream_inference_cost
0
provider_name
"DeepInfra"
Evaluation details
Result
Evaluator
Details
Meta Data
0.00%
Matches word count
n/a
neededClean
false
words
142
99.49%
Dialogue to Total Word Ratio
Ratio: 87.32%, Deviation: 2.68%
neededClean
false
wordsTotal
142
wordsDialogue
124
49.7442%