NC Bench
Overview
About
Tests
Models
Vendors
Run: 2026-02-20-01KHYAHRZ504R5QX180GF9JEDQ
Run Details
Model:
Gemma 3 12B
Test:
Dialogue tags
Scenario:
Write 500 words with 50% dialogue
Duration:
18.4s
Cost:
$0.000088
Provider:
DeepInfra
Evaluation summary
Score
Evaluator
Details Link
0.00%
Matches word count
(details)
0.00%
Dialogue to Total Word Ratio
(details)
0.0001%
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
The aroma of sugar cookies and desperation hung thick in Bill’s cramped apartment. Neon pink streamers, remnants of a forgotten birthday party, clung to the chipped wallpaper like stubborn glitter. Molly paced, a whirlwind of nervous energy in a leopard-print jumpsuit. Bill, meanwhile, was meticulously constructing a miniature replica of a bouncy castle from toothpicks and glue. “A bouncy castle, Bill? Seriously?” Molly finally stopped her pacing, hands on her hips. “We're about to rob a children’s party, and you're building a… a bouncy castle diorama?” “It’s for scale, Molly! Strategic planning!” Bill retorted, carefully affixing a tiny plastic slide. “Knowing the obstacles, the layout… it’s crucial. Plus,” he added with a small, mischievous grin, “it's kind of fun.” Molly rolled her eyes. “Fun? We’re talking about stealing presents from screaming five-year-olds! Do you have any idea the emotional trauma we're inflicting? We're practically villains!” “Technically, we’re opportunistic re-distributors of joy,” Bill corrected, not looking up from his construction. “Think of it as an intervention. Those kids have *way* too many toys.” “Right, and we’re precisely the people to correct that imbalance.” Molly sighed, running a hand through her brightly colored hair. "Okay, so let's go over the plan again. You distract the clown.” “I *am* the clown,” Bill said proudly, producing a floppy, oversized red nose from his pocket. “Already practicing my balloon animal skills. I’ll juggle, make ridiculous noises, generally be a nuisance until everyone's too distracted to notice us.” “And I’ll go for the presents,” Molly continued. “Priority is the Lego sets. Then the remote-controlled cars. And absolutely, *under no circumstances,* do you touch the pony. I saw pictures. That pony is terrifying.” “Agreed. The pony stares into your soul,” Bill shuddered, putting down the partially-built bouncy castle. “So, timing?” “The parents are all going to be lined up for the cake cutting. Prime opportunity. I’ll blend in, look like a slightly eccentric aunt. Lots of compliments about how adorable little Timmy is.” "Adorable Timmy," Bill repeated. “Right, and the exit strategy? Last time we almost got stuck in the bouncy house.” Molly grimaced. “We go through the back gate. I’ve memorized the route. There’s a rose bush, but I'll handle that. You just… don't get tangled in the balloons. Again.” “No promises,” Bill said, winking. He adjusted his red nose. "Alright, Molly. Let's go liberate some joy. And hopefully avoid the pony." "Just remember," Molly said, checking her watch. "Operation Candy Cane is a go at precisely 3:15. Don't be late, Bill, and for heaven’s sake, leave the miniature bouncy castle here!"
created_at
"2026-02-20T20:06:09.135Z"
2/20/2026, 8:06:09 PM
model
"google/gemma-3-12b-it"
app_id
182717
external_user
(null)
streamed
true
cancelled
false
latency
409
moderation_latency
(null)
generation_time
18286
tokens_prompt
93
tokens_completion
749
native_tokens_prompt
75
native_tokens_completion
654
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
0.00008802
router
(null)
provider_responses
0
endpoint_id
"eb06dc92-5a16-47ec-a776-6ef956457c47"
id
"chatcmpl-RY6hmnHoGXhkj3AekwZ3Ep2R"
is_byok
false
latency
49
model_permaslug
"google/gemma-3-12b-it"
provider_name
"DeepInfra"
status
200
api_type
"completions"
id
"gen-1771617969-kUgTzoa1QUd7LN1YIsmY"
upstream_id
"chatcmpl-RY6hmnHoGXhkj3AekwZ3Ep2R"
total_cost
0.00008802
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
431
0.00%
Dialogue to Total Word Ratio
Ratio: 66.59%, Deviation: 16.59%
neededClean
false
wordsTotal
431
wordsDialogue
287
0.0001%