NC Bench
Overview
About
Tests
Models
Model
MoonshotAI: Kimi K2.5
Test
Dialogue tags
Scenario
Write 500 words with 30% dialogue
Run ID
2026-02-08-01KGYYTSK138H9RRKDCX20YPQ1
Model input
User/Human
You are a author of comedic heist books. Please write a 500 word scene that has 30% 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
Molly adjusted her night vision goggles, which kept slipping because she’d bought the junior size to save twelve dollars. Through the chain-link fence, the backyard looked like a battlefield of primary colors. A bouncy castle quivered in the twilight, deflating like a dying jellyfish. Inside, sugar-crazed children screamed lyrics to something that sounded like “Baby Shark” but angrier. "Remember," Molly whispered, checking her utility belt, "we go in hard and fast. No lingering by the cake table this time." Bill was tangled in a garden hose. "I wasn't lingering. I was establishing a perimeter." "You asked the mom for the recipe." "Operational reconnaissance!" Molly consulted her clipboard, which had a house layout drawn in crayon because Bill had packed their blueprints with the decoy presents. "Entry point is the sliding glass door. Parents are distracted by the magician. We’ve got four minutes before the piñata." Bill freed himself. "What about the dog?" "Mr. Snuffles is sedated." "You didn't." "Benadryl meatball. He’s napping in the begonias." "Unethical." "We’re stealing from a five-year-old, Bill. Ethics left when you rented the getaway van with 'CLOWN SERVICES' on the side." "It was camouflage!" "It’s an Econoline with a teardrop painted under the headlight. We look like we process grief, not Legos." "Maybe we're grief clowns." "There's no such thing." "There could be." "Time?" "Thirty seconds." They crouched behind the recycling bin. A child wearing a pipe cleaner crown ran past, screaming about equity. The magician frantically chased doves away from the ceiling fan. "Go." They vaulted the fence—Bill got caught on a windmill—and landed in the grass. The presents sat on a picnic table wrapped in glitter paper. Molly spotted LEGO boxes, a Nintendo Switch, a dollhouse. Bill pulled out a potato sack. "This feels wrong." "It's a heist." "The sack itches." "Focus." They crept forward. The piñata swayed above the deck, a purple dinosaur judging them. "Wait for the distraction." "What distraction?" "The piñata." "Right. The donkey." "It's a dinosaur, Bill." Right on cue, the piñata exploded. Organic fruit leathers rained down; children swarmed like piranhas. Parents reached for wine. "Now." Bill sprinted, sack open. He grabbed a large box and turned. Then the alarm went off—a musical one. The birthday boy stepped on a piano mat, triggering "Twinkle Twinkle." Forty-seven heads turned. Bill held a Chemistry Set overhead, caught in LED lights. "Run?" "Run." They ran.
created_at
"2026-02-08T15:47:08.955039+00:00"
Sunday, February 8th, 2026 at 3:47:08 PM GMT+00:00
model
"moonshotai/kimi-k2.5-0127"
app_id
182717
external_user
(null)
streamed
true
cancelled
false
latency
805
moderation_latency
(null)
generation_time
133908
tokens_prompt
61
tokens_completion
6195
native_tokens_prompt
68
native_tokens_completion
6269
native_tokens_completion_images
(null)
native_tokens_reasoning
6307
native_tokens_cached
68
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
0.0188138
router
(null)
provider_responses
0
id
"chatcmpl-6988af760f3758ed7b2fe314"
status
200
is_byok
false
latency
805
endpoint_id
"49762299-66fd-4514-b0ee-c9e11e231e3b"
provider_name
"Moonshot AI"
model_permaslug
"moonshotai/kimi-k2.5-0127"
api_type
"completions"
id
"gen-1770565494-hVIjsEN4Cka6MGy8x02J"
upstream_id
"chatcmpl-6988af760f3758ed7b2fe314"
total_cost
0.0188138
cache_discount
0.000034
upstream_inference_cost
0
provider_name
"Moonshot AI"
Evaluation details
Result
Evaluator
Details
Meta Data
0.0000%
Matches word count
n/a
neededClean
false
words
402
0.8205%
Dialogue to Total Word Ratio
Ratio: 39.80%, Deviation: 9.80%
neededClean
false
wordsTotal
402
wordsDialogue
160
0.4103%