NC Bench
Overview
About
Tests
Categories
Models
Vendors
Compare
Run: 2026-09-02-01M1GVN8SZP7Y3KGN2GFJNT2H2
Run Details
Model:
Z.AI GLM 5.3 (Reasoning, Max)
Test:
Dialogue tags
Scenario:
Write 500 words with 30% dialogue
Duration:
2m 30s
Cost:
$0.037581
Provider:
Z.AI
Evaluation summary
Score
Evaluator
Details Link
35.00%
Matches word count
(details)
98.26%
Dialogue to Total Word Ratio
(details)
66.6304%
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 adjusted her binoculars in the driver's seat of the van, though the target was fifteen feet away and separated from her by nothing but a hedge trimmed into the shape of a duck. Behind the duck, six-year-olds swarmed a bouncy castle with the coordination of a failed riot. A clown made balloon animals for a boy who was crying about the balloon animals. "You're sure about the security?" Bill whispered. "It's a child's birthday party." "Rich people hire contractors." "Her father sells gutters." Bill studied the yard through a cardboard tube he called operational. "How long do we have?" "Presents at four. Cake after." "Then we have a window." "It's not a window. It's forty minutes and a sugar crash." They had been on stakeout for three hours, which was two hours and forty-five minutes longer than the actual theft would take. Bill had produced a schematic of the house on a napkin, complete with labeled entry points, a section marked PRESENT TABLE, and a drawing he insisted was a guard dog but which Molly had identified, correctly, as a stick figure. "Phase one," he said, tapping the napkin. "I pose as the entertainment and distract the parents." "You're not a clown." "I don't have to be good. I have to be clown-shaped." She let that go. Through the hedge, she watched the actual clown twist a balloon poodle with the solemnity of a bomb technician, only to have it rejected by the crying boy in favor of further crying. Somewhere beyond the bouncy castle sat the present table, a folding table sagging under wrapping paper, guarded by absolutely no one. "Phase two," Bill continued. "You move in low, sweep the presents, and we extract through the side gate." "The gate with the dog sign?" "Bluff. It's a schnauzer named Pockets." Molly lowered the binoculars and studied her partner of eleven years, a man who had once stolen a Monet and once, separately, spent two days trapped in a Las Vegas air vent, surviving on communion wafers. Something in his face was doing the thing it did when he genuinely believed in something stupid. "We fence an Easy-Bake Oven," she said. "That's the plan. That's the empire." "And a slime kit. Slime is hot right now." "The job is a crime against children." "The job is victimless. Kids lose presents constantly. Dogs eat them. We're basically a dog." At that moment a small girl in a paper crown approached the present table, straightened one bow with the care of a jeweler, and returned to the castle. Molly watched her go. Bill pretended not to, but his jaw did something guilty. "One present," Molly said. "We leave a better one and never speak of this again." Bill stared at the napkin. Then, with the gravity of a general surrendering, he folded it away. "I'll get the wrapping paper," he said. That afternoon they stole a hula hoop and replaced it with a robot the size of a dishwasher. Katie Piper told everyone for years that Santa came in July. Fernando, somehow, still ended up with the slime kit.
created_at
"2026-09-02T10:46:56.583Z"
9/2/2026, 10:46:56 AM
model
"z-ai/glm-5.3-20260816"
app_id
182717
external_user
(null)
streamed
true
cancelled
false
latency
918
moderation_latency
(null)
generation_time
150086
tokens_prompt
75
tokens_completion
7718
native_tokens_prompt
73
native_tokens_completion
8518
native_tokens_completion_images
(null)
native_tokens_reasoning
7852
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
0.0375814
router
(null)
user_agent
"langchainjs-openai/1.0.0 ((node/v24.19.0; linux; x64))"
http_referer
(null)
request_id
"req-1788346016-xKvjEjlnUTAXVR1GqqrU"
session_id
(null)
preset_id
(null)
api_type
"completions"
id
"gen-1788346016-LIymhEkROei9Z4ogV2DS"
upstream_id
"20260902184656dfb9835ab6fc4a93"
provider_responses
0
endpoint_id
"db975786-e7f6-4838-9e56-3e0b77f1bdc2"
id
"20260902184656dfb9835ab6fc4a93"
is_byok
false
latency
918
model_permaslug
"z-ai/glm-5.3-20260816"
provider_name
"Z.AI"
status
200
total_cost
0.0375814
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
35.00%
Matches word count
n/a
neededClean
false
words
518
98.26%
Dialogue to Total Word Ratio
Ratio: 30.59%, Deviation: 0.59%
neededClean
false
wordsTotal
523
wordsDialogue
160
66.6304%