NC Bench
Overview
About
Tests
Categories
Models
Vendors
Compare
Run: 2026-09-02-01M1H5J5GNYTABDB0D90RF5X1R
Run Details
Model:
Z.AI GLM 5.3 Flash (Reasoning, Low)
Test:
Dialogue tags
Scenario:
Write 500 words with 30% dialogue
Duration:
22.4s
Cost:
$0.000182
Provider:
Z.AI
Evaluation summary
Score
Evaluator
Details Link
27.17%
Matches word count
(details)
0.00%
Dialogue to Total Word Ratio
(details)
13.5829%
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 spread the blueprint across the trunk of the car, weighting the corners with a juice box, a rubber duck, and a bag of confetti she'd swiped from a previous job gone hilariously wrong. "Okay, here's the layout," she said, tapping the paper. "Bouncy castle in the backyard, ball pit in the garage, gift table right by the patio door." Bill squinted at the blueprint, then at the house they were crouched behind. "Why is there a unicorn drawn in the corner?" "That's not part of the layout. That's a mascot. It's going to be walking around. Do not make eye contact with the unicorn." "How am I supposed to not make eye contact with a six-foot unicorn, Molly?" "You look at its knees. Nobody checks a unicorn's knees." Bill rubbed his temples. "Remind me why we're stealing presents from children again." "Because the birthday boy's father is Teddy Kowalski, money launderer extraordinaire, and rumor is he launders more than money in those gift boxes. Gift cards, bearer bonds, possibly the crown situation from that museum job. The kid gets more presents than a small nation." "So we rob a drug-adjacent clown party." "We rob a drug-adjacent clown party very carefully." Bill studied the yard. "There are forty children out there." "Thirty-eight. Two are in the ball pit. Ball pit's our cover. Kids in a ball pit see an adult, they assume it's a parent. They don't question anything. I once walked out of a birthday party with an entire tricycle because a kid in a ball pit saw me and yelled 'bye, typhoon uncle.'" "That's not a real thing kids say." "That's exactly what Marcus yelled. Typhoon uncle. I've never been prouder of a cover identity I didn't even create." Bill pulled out his binoculars. "Okay, gift table. Sixteen presents, minimum. The big one has a ribbon the size of a car air freshener. That's either the bonds or a pony." "If it's a pony, we abort." "We abort if it's a pony?" "Molly's Heist Rule Nine: you do not steal a child's pony. There are lines, Bill." "You have a rule numbered nine? What are the other eight?" "Mostly about the unicorn." Bill lowered the binoculars. "The unicorn isn't even that big a deal—" A costumed figure rounded the corner of the house, glitter horn first. Bill dropped the binoculars and fell backward into a bush. "Knees, Bill," Molly hissed. "Look at its knees." "It's eight feet tall and it's shimmering, Molly!" "Then look at its shimmering knees." The unicorn paused, seemed to peer in their direction, then bounced away toward the bouncy castle. Bill emerged from the bush with a twig in his hair. "I hate this job. I used to steal cars. Cars don't shimmer." "In and out in four minutes. I take the table, you cut the fence alarm. We're out before the cake." "And if it's a pony?" Molly grinned and pulled on her gloves. "Then we leave a very apologetic note and steal the clown's wallet instead." "I hate the clown's wallet." "Everybody hates the clown's wallet. That's why it's Rule Eight."
created_at
"2026-09-02T13:40:00.679Z"
9/2/2026, 1:40:00 PM
model
"z-ai/glm-5.3-flash-20260826"
app_id
182717
external_user
(null)
streamed
true
cancelled
false
latency
1532
moderation_latency
(null)
generation_time
22338
tokens_prompt
75
tokens_completion
771
native_tokens_prompt
73
native_tokens_completion
723
native_tokens_completion_images
(null)
native_tokens_reasoning
10
native_tokens_cached
64
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
0.000182385
router
(null)
user_agent
"langchainjs-openai/1.0.0 ((node/v24.19.0; linux; x64))"
http_referer
(null)
request_id
"req-1788356400-pBtXNAfzo9LGicoTZQxz"
session_id
(null)
preset_id
(null)
api_type
"completions"
id
"gen-1788356400-po7CifWVVxIC2zwX8NqA"
upstream_id
"20260902214000be6f74ca5bb44285"
provider_responses
0
endpoint_id
"8e9fe48b-2f91-41c3-a8a7-e4a93a8c4ff0"
id
"20260902214000be6f74ca5bb44285"
is_byok
false
latency
1531
model_permaslug
"z-ai/glm-5.3-flash-20260826"
provider_name
"Z.AI"
status
200
total_cost
0.000182385
cache_discount
0.00000768
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
27.17%
Matches word count
n/a
neededClean
false
words
519
0.00%
Dialogue to Total Word Ratio
Ratio: 75.72%, Deviation: 45.72%
neededClean
false
wordsTotal
523
wordsDialogue
396
13.5829%