NC Bench
Overview
About
Tests
Categories
Models
Vendors
Compare
Run: 2026-09-02-01M1GTMWDQEPS7NYFP599YC70T
Run Details
Model:
Z.AI GLM 5.3 (Reasoning, Max)
Test:
Dialogue tags
Scenario:
Write 500 words with 30% dialogue
Duration:
4m 11s
Cost:
$0.067537
Provider:
Z.AI
Evaluation summary
Score
Evaluator
Details Link
98.71%
Matches word count
(details)
97.41%
Dialogue to Total Word Ratio
(details)
98.0625%
Overall
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 crouched behind the hydrangeas, binoculars pressed to her face despite the target being plainly visible to anyone with functioning eyeballs. Three houses down, the Henderson backyard was a festival of primary colors: a bouncy castle sagging in the heat, a unicorn piñata strung from the oak, and a folding table heaped with presents in wrapping paper so bright it could be seen from space, or at least from a van with tinted windows. Bill lay beside her in the grass, wearing a Hawaiian shirt he insisted qualified as civilian camouflage. He was eating trail mix from home, which Molly considered unprofessional on at least four levels. "The perimeter's soft," Molly said, scanning. "No cameras. No dog. One clown." "Clowns are variables," Bill said through a mouthful of almonds. "You don't plan around a clown. You pray." "It's a birthday clown, Bill. His only weapon is a balloon poodle." "I've been hit by a balloon poodle. Cleveland. 2019. My ear never recovered." Molly lowered the binoculars and looked at him the way she often did — like a woman recalculating her retirement in real time. The present table gleamed in the afternoon sun like a shrine to untraceable resale value. "Our window is three-fifteen, when the cake comes out. Every parent in that yard will point a phone at one candle. Sixty seconds of blindness. We go in low, empty the table, and we're gone before the kid finishes wishing." "Which, statistically, is a wish for a puppy. Just saying." The plan was, on paper, flawless. It was drafted on a Denny's napkin, a step up from their previous blueprint, which lived on the back of a speeding ticket. They were professionals. They had stolen a Monet from a gallery in Nice, a Fabergé egg from a Monaco sheikh, and, through a series of misunderstandings, a zoo. A children's party in the suburbs should have been beneath them. That was exactly why Molly loved it. "No fences," she explained, watching the clown twist a balloon into something unlawful. "No buyers, no auction houses, no man named Sergei asking about your mother's health. LEGOs are untraceable. They're the new bearer bonds. Every kid in this county wants the same three sets." Bill considered this, chewing. "So we're fencing LEGOs." "We're redistributing inventory." Beyond the fence, a child shrieked with the pure joy of a sugar rush, and Molly checked her watch. Three-fourteen. The lawn was about to become an ocean of distracted adults filming vertically. She pulled a black beanie on despite the ninety-degree heat, because a heist had standards. "Sixty seconds," she whispered. "Presents only." "Obviously," said Bill, whose eyes had drifted to the unicorn piñata like a compass finding north. "Presents only." "Bill." "Thirty-five seconds is plenty to grab a piñata." "Bill, I swear—" But the singing had started — off-key, deafening, twenty children strong — and the window was open. Molly went over the fence like a shadow. Bill followed like a man who had just, quietly, revised the mission parameters, and whose pockets were already unzipped.
created_at
"2026-09-02T10:29:15.331Z"
9/2/2026, 10:29:15 AM
model
"z-ai/glm-5.3-20260816"
app_id
182717
external_user
(null)
streamed
true
cancelled
false
latency
2801
moderation_latency
(null)
generation_time
250637
tokens_prompt
75
tokens_completion
12976
native_tokens_prompt
73
native_tokens_completion
15326
native_tokens_completion_images
(null)
native_tokens_reasoning
14631
native_tokens_cached
0
num_media_prompt
(null)
num_input_audio_prompt
(null)
num_media_completion
0
num_search_results
(null)
num_fetches
(null)
web_search_engine
(null)
origin
"https://ncbench.com/"
is_byok
false
finish_reason
"stop"
native_finish_reason
"stop"
service_tier
(null)
usage
0.0675366
router
(null)
user_agent
"langchainjs-openai/1.0.0 ((node/v24.19.0; linux; x64))"
http_referer
(null)
request_id
"req-1788344955-kJObcxnRWrnFRzQAtv9W"
session_id
(null)
preset_id
(null)
api_type
"completions"
id
"gen-1788344955-do0yLxqH182KuhTRFiP9"
upstream_id
"20260902182915aa96196c36aa4b66"
provider_responses
0
endpoint_id
"db975786-e7f6-4838-9e56-3e0b77f1bdc2"
id
"20260902182915aa96196c36aa4b66"
is_byok
false
latency
2801
model_permaslug
"z-ai/glm-5.3-20260816"
provider_name
"Z.AI"
status
200
total_cost
0.0675366
cache_discount
(null)
upstream_inference_cost
0
provider_name
"Z.AI"
response_cache_source_id
(null)
data_region
"global"
workspace_id
"97e315e5-d303-487d-83c1-83180e8a13d4"
Evaluation details
Result
Evaluator
Details
Meta Data
98.71%
Matches word count
n/a
neededClean
false
words
506
97.41%
Dialogue to Total Word Ratio
Ratio: 30.72%, Deviation: 0.72%
neededClean
false
wordsTotal
511
wordsDialogue
157
98.0625%