NC Bench
Overview
About
Tests
Models
Model
Z.AI GLM 4.7
Test
Dialogue tags
Scenario
Write 500 words with 50% dialogue
Run ID
2026-02-08-01KGYXY2N2EM6N3SJM31W62VKK
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
The beige station wagon in the driveway mocked Bill’s very existence. He crouched behind the oleander bush, trying to scratch his nose without disturbing the pollen that was currently attempting to suffocate him. "Target is soft," Molly whispered, adjusting her tactical balaclava. "The perimeter is compromised by streamers." "It’s a five-year-old’s birthday, Molls, not Fort Knox," Bill grunted, shifting his weight off a sharp rock. "Can I take the mask off? It’s ninety degrees out here." "Negative. Visibility is key. We maintain cover until the Musical Chairs distraction initiates," she stated, checking her watch. "Do you have the heavy-duty trash bags?" "Right here," Bill patted his cargo pocket. "But are we really doing this? Stealing Lego sets? My criminal record is for grand larceny. This feels like a step down." "Think about the resale value on the black market," Molly countered, eyes narrowing behind her sunglasses. "The limited edition action figures alone are worth five grand. We hit the pile, we grab the loot, and vanish before the candle lighting." "The kids are going to cry," Bill noted softly. "Collateral damage," she dismissed with a wave of her gloved hand. "Look, the team is assembling. The dads are forming a defensive line around the cooler. We need to bypass the main force." "How do we bypass a dozen guys drinking light beer?" "We blend," Molly said, pulling two rubber noses from her utility belt. "Put this on. We walk in like eccentric uncles. We grab the gift mountain, and walk out." "A rubber nose? That’s the disguise?" "It’s psychological warfare, Bill. No one suspects a guy with a rubber nose," she insisted. "On my mark. Three, two, one..." "Wait, is that a piñata?" Bill panicked, eyeing the colorful donkey hanging from the oak tree. "Hold position! It’s a decoy!" "It’s shaped like a donkey, Molls!" "It’s a trap! Stay down!" she grabbed his shoulder. "We wait for the sugar crash. The parents will be weak. The children will be delirious. That is our window." "I hate you so much right now," Bill groaned, attaching the red sphere to his face. "Save the passion for the getaway driver," Molly quipped, crouching lower. "Remember, we leave no wrapping paper behind. We’re professionals." "I’ll be sure to use a dustpan," he muttered. "Silence. The clown is parking."
created_at
"2026-02-08T15:31:46.718866+00:00"
Sunday, February 8th, 2026 at 3:31:46 PM GMT+00:00
model
"z-ai/glm-4.7-20251222"
app_id
182717
external_user
(null)
streamed
true
cancelled
false
latency
1595
moderation_latency
(null)
generation_time
152387
tokens_prompt
61
tokens_completion
2247
native_tokens_prompt
66
native_tokens_completion
2270
native_tokens_completion_images
(null)
native_tokens_reasoning
1880
native_tokens_cached
64
num_media_prompt
(null)
num_input_audio_prompt
(null)
num_media_completion
0
num_search_results
(null)
origin
"https://ncbench.com/"
is_byok
false
finish_reason
"stop"
native_finish_reason
"stop"
usage
0.00397842
router
(null)
provider_responses
0
id
"chatcmpl-RAhRm40QeLhwr24aQVqdEBNS"
status
200
is_byok
false
latency
64
endpoint_id
"ab79ffa8-8dcc-41c4-9a9f-6cc408135243"
provider_name
"DeepInfra"
model_permaslug
"z-ai/glm-4.7-20251222"
api_type
"completions"
id
"gen-1770564553-IyGFECj7O48mn54MB0BD"
upstream_id
"chatcmpl-RAhRm40QeLhwr24aQVqdEBNS"
total_cost
0.00397842
cache_discount
0.00002048
upstream_inference_cost
0
provider_name
"DeepInfra"
Evaluation details
Result
Evaluator
Details
Meta Data
0.0000%
Matches word count
n/a
neededClean
false
words
387
0.0000%
Dialogue to Total Word Ratio
Ratio: 69.25%, Deviation: 19.25%
neededClean
false
wordsTotal
387
wordsDialogue
268
0.0000%