NC Bench
Overview
About
Tests
Models
Vendors
Run: 2026-02-20-01KHY9D6ZQ4GF2FY6Q5A3ESCJN
Run Details
Model:
Gemma 3 4B
Test:
Dialogue tags
Scenario:
Write 500 words with 50% dialogue
Duration:
8.4s
Cost:
$0.000060
Provider:
DeepInfra
Evaluation summary
Score
Evaluator
Details Link
0.00%
Matches word count
(details)
0.00%
Dialogue to Total Word Ratio
(details)
0.0000%
Overall
Model input
User/Human
You are a author of comedic heist books. Please write a 500 word scene that has 50% 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. Obvious entry point, wouldn't you say?” Molly asked, meticulously arranging miniature plastic dinosaurs on a napkin. “A bouncy castle? Really, Molly? We’re going for the bouncy castle?” Bill squinted, pushing a stray lock of grey hair back from his forehead. “Feels a little… enthusiastic. Like we're infiltrating a preschool.” “Enthusiastic is good, Bill! It’s festive. It’s important. Think of the chaos! The sheer, unadulterated joy of a coordinated present snatching while a five-year-old shrieks about a glitter bomb.” “Glitter bomb? You’ve lowered the stakes significantly. I was picturing something… more sophisticated. Like a ventilation shaft. Or a disgruntled butler.” “The ventilation shaft is a death trap. And Mr. Henderson, the butler, is currently napping with a plate of digestive biscuits. He’d probably just snore through the whole thing.” “Point taken. So, bouncy castle it is. How do we get in?” “Simple. I’ve procured a very convincing inflatable unicorn costume. Think pink, think sparkly, think 'slightly terrifying adult attempting to look like a child’s dream.’” Bill sputtered. “A unicorn? You’re seriously suggesting a unicorn?” “It’s our cover, Bill! No one suspects a unicorn. Plus, I can squeeze through gaps. It’s surprisingly flexible, this fabric.” Molly patted the costume draped over a chair. “Think of it as a mobility enhancement.” “I’m thinking of it as a public humiliation waiting to happen. What about the parents? They’ll be watching.” “They’ll be too busy supervising sugar rushes and preventing meltdowns to notice a slightly odd adult in a unicorn suit. Besides, I’ve programmed a loop of baby shark on my smartwatch. It’s surprisingly distracting.” “Baby shark? You’re actively trying to make this worse.” “It’s a tactical sound weapon, Bill! Honestly, you're being difficult. Okay, let’s run through the plan. I go in, unicorn disguise, secure the perimeter – mostly the cupcake table. You handle the presents. Specifically, anything with a cartoon character. Priority one.” “What about the sheer volume of presents? We’re talking a veritable mountain of plastic figurines and remote controls.” “We’ll prioritize. Think plushies. Think Lego. Think… any item that screams ‘child’s obsession.’” Molly paused, tapping the napkin thoughtfully. “I’ve also identified a blind spot near the magician’s table. He’s impressively bored, seems like a good distraction point.” “A bored magician? You're terrifyingly prepared. Look, the whole thing just… it feels wrong, Molly. Stealing presents. It’s…tacky.” “Tacky is fun, Bill! Tacky is lucrative! And frankly, those parents need to learn a lesson about over-gifting. Now, seriously, are you in or are you going to spend the rest of the afternoon debating the ethical implications of unicorn-based heist operations?” “Fine, fine. Unicorn it is. But if I’m stuck in a bouncy castle arguing with a five-year-old about the merits of sparkly stickers, I’m blaming you.”
created_at
"2026-02-20T19:46:11.075Z"
2/20/2026, 7:46:11 PM
model
"google/gemma-3-4b-it"
app_id
182717
external_user
(null)
streamed
true
cancelled
false
latency
193
moderation_latency
(null)
generation_time
7672
tokens_prompt
93
tokens_completion
834
native_tokens_prompt
75
native_tokens_completion
708
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
0.00005964
router
(null)
provider_responses
0
endpoint_id
"d3de3bd4-81bc-48fb-924f-2a87b2a36e75"
id
"chatcmpl-RtjsIMCaGlsILGeiLPvi8fKS"
is_byok
false
latency
51
model_permaslug
"google/gemma-3-4b-it"
provider_name
"DeepInfra"
status
200
api_type
"completions"
id
"gen-1771616771-sveHwzY6Ivt1TPryhUGQ"
upstream_id
"chatcmpl-RtjsIMCaGlsILGeiLPvi8fKS"
total_cost
0.00005964
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
462
0.00%
Dialogue to Total Word Ratio
Ratio: 91.56%, Deviation: 41.56%
neededClean
false
wordsTotal
462
wordsDialogue
423
0.0000%