NC Bench
Overview
About
Tests
Categories
Models
Vendors
Compare
Run: 2026-09-02-01M1GTZHW3KAH6VVR97SDZKZVN
Run Details
Model:
Z.AI GLM 5.3 (Reasoning, Max)
Test:
Dialogue tags
Scenario:
Write 500 words with 50% dialogue
Duration:
8m 53s
Cost:
$0.150763
Provider:
Z.AI
Evaluation summary
Score
Evaluator
Details Link
99.98%
Matches word count
(details)
98.77%
Dialogue to Total Word Ratio
(details)
99.3782%
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's kitchen had been converted into a command center, meaning the refrigerator was papered with crayon drawings of the Maple Street Church community hall and a casserole dish was pinning down what she insisted was a blueprint. Bill had dressed for a briefing and assumed there would be more lamination involved. "Target: Tommy Henderson's seventh birthday. Saturday, two o'clock. Presents here. Bouncy castle blocking the east exit," she said, tapping the paper. Bill squinted at the blueprint, then held it up to the light, as if that might reveal professional drafting. "This is a drawing. You did this in crayon." "Recon costs money, Bill. I work with what I have." He tapped a purple blob with the gravity of a man reviewing satellite imagery. "What's this?" "The bouncy castle. You can't sprint across one. Physics won't allow it." "Then we don't sprint across it." "Correct. We go in as the entertainment. I booked us as magicians." Bill went pale in the specific way of a man watching a traumatic memory clear its throat. "You made me play a chimney sweep and sold me out to a real one." "One time. Besides, magicians get full access to the presents. Two hundred dollars, too." "Magicians are the least trusted people at a kid's party. There's always one checking your dove for the trapdoor." "His name is Dylan. I watched him interrogate a pizza man for eleven minutes. Avoid Dylan." Bill leaned back and rubbed his eyes with the weary air of a man who'd asked before. "How do we get twenty presents past forty parents, a pastor, and a grandmother knitting what looks like a garrote?" Molly uncapped a marker and drew a triumphant circle around a stick figure dangling from a tree. "The piƱata. When it breaks, every kid lunges for candy. Every parent lunges after them. The pastor lunges because he's polite. Ninety-second window." "You've timed this." "The Ricci party. Research. During the chaos, I load the wagon and you cover me." "Cover me how?" "You cry." The marker squeaked to a stop. Somewhere upstairs a television laughed at something. Bill stared at her the way a man stares at a slot machine that has just swallowed his wedding ring. "A grown magician weeping at a party confuses everyone. Nobody interrogates a crying man, Bill. Nobody." "What if the kids cry?" Molly paused, the marker hovering over the blueprint. Outside, a lawnmower droned on, indifferent to crime. "The birthday boy keeps his presents. His grandmother's, too. I'm a thief, Bill, not a monster." Bill picked up a stray crayon and twirled it, weighing his soul against two hundred dollars and a wagon of stolen toys. Then, without warning, he contorted his face into something tragic and glistening. "What are you doing?" "Practicing my crying face. Is it working?" Molly studied him for a long moment with the critical eye of a former county fair pie judge. "It's genuinely disturbing," she said. "The parents will love it." Somewhere upstairs, the television laughed again.
created_at
"2026-09-02T10:35:04.971Z"
9/2/2026, 10:35:04 AM
model
"z-ai/glm-5.3-20260816"
app_id
182717
external_user
(null)
streamed
true
cancelled
false
latency
2961
moderation_latency
(null)
generation_time
532883
tokens_prompt
75
tokens_completion
27337
native_tokens_prompt
73
native_tokens_completion
34241
native_tokens_completion_images
(null)
native_tokens_reasoning
33568
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
0.1507626
router
(null)
user_agent
"langchainjs-openai/1.0.0 ((node/v24.19.0; linux; x64))"
http_referer
(null)
request_id
"req-1788345304-wdgoonD8VaIWVIIY6T0y"
session_id
(null)
preset_id
(null)
api_type
"completions"
id
"gen-1788345304-v9Xe1VnbHAObnsR2zXCO"
upstream_id
"20260902183505d4ed9b0003594667"
provider_responses
0
endpoint_id
"db975786-e7f6-4838-9e56-3e0b77f1bdc2"
id
"20260902183505d4ed9b0003594667"
is_byok
false
latency
2961
model_permaslug
"z-ai/glm-5.3-20260816"
provider_name
"Z.AI"
status
200
total_cost
0.1507626
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
99.98%
Matches word count
n/a
neededClean
false
words
502
98.77%
Dialogue to Total Word Ratio
Ratio: 50.50%, Deviation: 0.50%
neededClean
false
wordsTotal
503
wordsDialogue
254
99.3782%