NC Bench
Overview
About
Tests
Models
Vendors
Run: 2026-02-20-01KHY7XJDGHG7DWVSG2RGR26DQ
Run Details
Model:
Gemma 3 27B
Test:
Dialogue tags
Scenario:
Write 500 words with 70% dialogue
Duration:
25.1s
Cost:
$0.000146
Provider:
Novita
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 70% 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 recap. Rainbows, glitter, small humans fueled by sugar. This is our target environment?” Bill asked, meticulously polishing his glasses with a handkerchief. Molly rolled her eyes. “You’ve said ‘small humans fueled by sugar’ five times, Bill. Yes. It’s a six-year-old’s birthday party. And it’s a goldmine. Apparently little Timmy’s parents went *overboard*.” “Overboard with the presents, or overboard with the… bouncy castle security?” “Both, probably. Which is why we need a plan. A *good* plan. Not your ‘blend in, compliment the cake’ plan.” Bill bristled. “That’s a solid plan! People love cake compliments. It disarms them.” “It gets you identified as the weird adult hovering around the snacks, Bill. We’re going for subtle. I’m thinking distraction. You, my friend, are going to be a magician.” “A magician? Molly, I can barely tie my own shoelaces, let alone pull a rabbit—" “Not a real magician. A… convincing one. I’ve got a kit. Smoke pellets, a fake dove, a deck of cards. You wave things around, say some gibberish, and everyone looks at the sparkly distractions while *I* relieve Timmy of his loot.” “Sparkly distractions? Is that what we’re calling petty theft these days?” “It’s strategic acquisition! And it's sparkly. Look, you just need to keep them occupied for, say, five minutes. Five minutes while I empty the present table into these oversized tote bags.” Molly brandished two floral shopping bags. Bill stared at them. “Those look like my Aunt Mildred’s grocery bags.” “They were the least conspicuous! What did you expect, black duffel bags labeled ‘STOLEN PRESENTS’?” “A valid point.” He sighed. “Okay, magician. But what about the parents? They’re going to notice a mountain of gifts suddenly… dwindling.” “That’s where the chaos comes in. Confetti cannons. I’m setting them off as cover. Plus, I’ve heard Timmy’s grandmother has a reputation for being… enthusiastic during the limbo competition. We unleash Grandma Betty and the confetti, and everyone’s too busy dodging low-hanging bars and rainbow paper to notice a couple of slightly-too-old party guests making off with a Playstation 5.” “Grandma Betty? You’ve been doing your research.” “Always, Bill. Always. Now, practice your ‘abracadabra’. And try not to set the bouncy castle on fire.” “No promises. This whole thing feels… morally ambiguous, Molly.” “We’re stealing from a six-year-old, Bill. It’s *extremely* morally ambiguous. But think of it as… redistributing wealth.” “To whom? Us?” “To… people who appreciate fine porcelain dolls and limited edition action figures! It’s for the greater good.” Bill adjusted his glasses. “Right. For the greater good. As long as Aunt Mildred doesn’t ask where I got her grocery bags.”
created_at
"2026-02-20T19:20:09.921Z"
2/20/2026, 7:20:09 PM
model
"google/gemma-3-27b-it"
app_id
182717
external_user
(null)
streamed
true
cancelled
false
latency
732
moderation_latency
(null)
generation_time
25040
tokens_prompt
93
tokens_completion
783
native_tokens_prompt
74
native_tokens_completion
688
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
0.000146406
router
(null)
provider_responses
0
endpoint_id
"cf530f36-7cb8-4eb9-8953-880eef952af3"
id
"4e1b3325-366b-4881-b3b8-6d6036178f10"
is_byok
false
latency
620
model_permaslug
"google/gemma-3-27b-it"
provider_name
"Novita"
status
200
api_type
"completions"
id
"gen-1771615209-rLOtDhcXqK2MdWsXZTpf"
upstream_id
"4e1b3325-366b-4881-b3b8-6d6036178f10"
total_cost
0.000146406
cache_discount
(null)
upstream_inference_cost
0
provider_name
"Novita"
Evaluation details
Result
Evaluator
Details
Meta Data
0.00%
Matches word count
n/a
neededClean
false
words
439
0.00%
Dialogue to Total Word Ratio
Ratio: 92.94%, Deviation: 22.94%
neededClean
false
wordsTotal
439
wordsDialogue
408
0.0000%