NC Bench
Overview
About
Tests
Categories
Models
Vendors
Compare
Run: 2026-09-02-01M1GQZRS1GSK9Y5YX2KZD72DT
Run Details
Model:
Z.AI GLM 5.3 Flash (Reasoning, Max)
Test:
Dialogue tags
Scenario:
Write 500 words with 50% dialogue
Duration:
2m 57s
Cost:
$0.002247
Provider:
Z.AI
Evaluation summary
Score
Evaluator
Details Link
0.03%
Matches word count
(details)
82.48%
Dialogue to Total Word Ratio
(details)
41.2560%
Overall
Model input
User/Human
You are a author of comedic heist books. Please write a 500 word scene that has 50% 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 spread the blueprint across the card table, weighting the corners with whatever the safehouse had to offer: an ashtray, a coffee mug, and a rubber duck that had been staring at Bill all week. The blueprint itself was drawn in crayon, which Bill had assumed was a joke until he realized Molly had stolen it from the victim. She tapped the drawing. "Target is the Henderson kid's seventh birthday. Saturday, two p.m., backyard of 44 Maple Drive." "Security?" Bill studied the map like it was Fort Knox. "One mom with a clipboard. Bouncy castle in the east quadrant. And a clown, but he's freelance, so don't expect loyalty if things go sideways." Bill whistled at the haul scrawled in the margin: a bike, a gaming console, and something called a Polly Pocket Dance Studio. "That's a big score for a seven-year-old," he said. "He won't miss it. He'll cry for twenty minutes, then he'll get cake. Kids are resilient. It's why we target them." Bill rubbed his jaw. He'd been in the business long enough to know when a plan was insane and when it was merely deranged, and this one was doing an impressive impression of both. "We can't just walk into a yard full of parents." "Which is why we don't walk in. We go under." Molly tapped the storm drain that ran from the creek to the petunias. "We surface, grab the present pile during musical chairs, and slide back in before anyone notices." "Musical chairs is loud. Everyone watches the chairs." "Everyone watches the chairs, Bill. Nobody watches the presents." He conceded the point with a grunt, which was as close to agreement as he ever came before his second coffee. "What about the clown?" "Balloon animals by the fence. Nowhere near the pile." "I don't do clowns, Molly. Seoul, 2019—" "You were the clown in Seoul, Bill." "That clown had a code." "The clown got us chased through a mall by a security guard on a Segway." Bill folded his arms, a gesture that had once intimidated bank managers and now intimidated no one. "Fine. But if this clown starts anything, I finish it." "Nobody's finishing a clown. You're on gift retrieval, I'm on lookout." "What's my cover?" Molly slid a purple t-shirt with a cartoon dinosaur across the table. He held it up against himself and sighed, because it fit, and because deep down he'd known it would. "I look ridiculous," he said. "You look forgettable. That's the whole game." "And the getaway?" "Pink bicycle, basket, streamers. We ride out like we're late for nap time." "A bicycle with streamers," Bill repeated, testing the words for a trap. "Where did you even get it?" "Stole it from a nine-year-old last week. Call it research." Bill decided not to ask whether she meant the bicycle or the technique. He stared at the blueprint for a long moment, and somewhere in the walls the radiator knocked twice, like it was laughing. "There's a bike in the take, you said." "There is." "Twenty-six inch wheels?" "Twenty." "That's insulting, Molly. I'm a professional. I have standards." "You're a grown man stealing from children, Bill. Adjust your standards."
created_at
"2026-09-02T09:42:46.316Z"
9/2/2026, 9:42:46 AM
model
"z-ai/glm-5.3-flash-20260826"
app_id
182717
external_user
(null)
streamed
true
cancelled
false
latency
1623
moderation_latency
(null)
generation_time
176578
tokens_prompt
75
tokens_completion
8146
native_tokens_prompt
73
native_tokens_completion
8967
native_tokens_completion_images
(null)
native_tokens_reasoning
8260
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
0.002247225
router
(null)
user_agent
"langchainjs-openai/1.0.0 ((node/v24.19.0; linux; x64))"
http_referer
(null)
request_id
"req-1788342166-cuwjlBUF9iwW8XjW16R4"
session_id
(null)
preset_id
(null)
api_type
"completions"
id
"gen-1788342166-z38DJkv1ls4Jq4zJXF8x"
upstream_id
"202609021742463e6b78ca8af44490"
provider_responses
0
endpoint_id
"8e9fe48b-2f91-41c3-a8a7-e4a93a8c4ff0"
id
"202609021742463e6b78ca8af44490"
is_byok
false
latency
1622
model_permaslug
"z-ai/glm-5.3-flash-20260826"
provider_name
"Z.AI"
status
200
total_cost
0.002247225
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
0.03%
Matches word count
n/a
neededClean
false
words
530
82.48%
Dialogue to Total Word Ratio
Ratio: 51.96%, Deviation: 1.96%
neededClean
false
wordsTotal
535
wordsDialogue
278
41.2560%