NC Bench
Overview
About
Tests
Models
Model
Z.AI GLM 4.6
Test
Dialogue tags
Scenario
Write 500 words with 30% dialogue
Run ID
2026-02-08-01KGYYQ3B50D3WPZ75KXXRRCYN
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
The inside of the van smelled faintly of stale fries and desperation. A single, bare bulb swung from the ceiling, casting a jaundiced light on a map of suburban paradise that was actually a drawing on the back of a pizza box. Molly, her blonde hair pulled into a tight bun that screamed ‘military librarian,’ stabbed a finger at a crude rectangle labeled ‘PRESENTS.’ “The primary objective is the gift table, located adjacent to the bouncy castle. We call it ‘The Bounce Zone.’” Bill, slouched in the passenger seat, noisily crunched a pork rind. “You call it ‘The Bounce Zone’? I’m calling it ‘The Vomit Gauntlet.’ Have you seen a six-year-old after three slices of cake and a juice box? It’s a biohazard waiting to happen.” Molly ignored him, her eyes tracing the perimeter she’d drawn in red crayon. “Distraction is key. Bill, you’re on clown duty. Jingles. He’s the wild card.” “What am I supposed to do, challenge him to a balloon animal-off? I can’t make a poodle. My giraffes look like deflated snakes with glandular issues.” “You won’t have to,” she said, her voice sharp. “At 14:00 hours, you ‘accidentally’ release the piñata. That will draw all assets—children and parents—to the back yard. It creates a ten-minute window.” Bill wiped his greasy fingers on his jeans. “And you’ll grab the presents?” “I’ll initiate the extraction. The biggest risk is auditory. The crinkle of wrapping paper is a dead giveaway. We need duffel bags. Lots of duffel bags.” He shuddered, thinking of the brightly colored parcels. “What if we get a doll that cries? Or one of those dinosaurs that roars? Our cover’s blown, Molls. We’ll be trampled by a horde of screaming seven-year-olds wielding tiny plastic swords.” Molly finally looked up, a glint of manic determination in her eyes. “Then we don’t get talking toys. We only grab the quiet ones. It’s called asset selection, Bill. Try to keep up.”
created_at
"2026-02-08T15:43:59.528944+00:00"
Sunday, February 8th, 2026 at 3:43:59 PM GMT+00:00
model
"z-ai/glm-4.6"
app_id
182717
external_user
(null)
streamed
true
cancelled
false
latency
768
moderation_latency
(null)
generation_time
65583
tokens_prompt
61
tokens_completion
1764
native_tokens_prompt
66
native_tokens_completion
1772
native_tokens_completion_images
(null)
native_tokens_reasoning
1336
native_tokens_cached
64
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
0.0030892599999552
router
(null)
provider_responses
0
id
"chatcmpl-Rg4mDn4bwSKoB96sL6HEd9XT"
status
200
is_byok
false
latency
59
endpoint_id
"c8607edc-8577-4199-b03d-1213da4743dc"
provider_name
"DeepInfra"
model_permaslug
"z-ai/glm-4.6"
api_type
"completions"
id
"gen-1770565373-wl0XidcmvqfeShY1NsD2"
upstream_id
"chatcmpl-Rg4mDn4bwSKoB96sL6HEd9XT"
total_cost
0.0030892599999552
cache_discount
0.0000224000000448
upstream_inference_cost
0
provider_name
"DeepInfra"
Evaluation details
Result
Evaluator
Details
Meta Data
0.0000%
Matches word count
n/a
neededClean
false
words
330
0.0000%
Dialogue to Total Word Ratio
Ratio: 62.12%, Deviation: 32.12%
neededClean
false
wordsTotal
330
wordsDialogue
205
0.0000%