NC Bench
Overview
About
Tests
Categories
Models
Vendors
Compare
Run: 2026-02-25-01KJBE190S33GF1VGC98NWMTMH
Run Details
Model:
Qwen 3.5 27B
Test:
Dialogue tags
Scenario:
Write 500 words with 50% dialogue
Duration:
1m 16s
Cost:
$0.018104
Provider:
Alibaba
Evaluation summary
Score
Evaluator
Details Link
0.00%
Matches word count
(details)
71.69%
Dialogue to Total Word Ratio
(details)
35.8432%
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 adjusted her night-vision goggles. They pinched her nose. Bill sat in the passenger seat. He chewed loudly on gummy bears. The van idled quietly in the dark driveway. Outside, the party house pulsed with neon light. Children screamed inside. The engine hummed a low note. "So, the plan is simple," Molly said. "We get in, we grab the loot, we leave." Bill swallowed a bear. Crumbs fell on his tactical vest. He looked at the window. "Which loot are we targeting specifically?" Bill asked. "Are we talking about the cake?" "No, presents," Molly snapped. "Under the tree. All of them." Bill swallowed hard. He wiped his hands on his pants. "Why?" Bill asked. "Usually we hit banks." "Banks have lasers," Molly explained. "This place has a clown. It is the perfect distraction." "A clown?" Bill raised an eyebrow. "Is he armed?" "With balloons," Molly said. "Check your gear. Do you have the grabber?" Bill patted his pockets. He found the metal pole. "I got the grabber," Bill said. "I got the bag. I got the bears." "Good," Molly nodded. "Remember, no breaking toys. We need them for resale." "Resale?" Bill stopped chewing. "Who buys stolen Legos?" "People who lost their own," Molly replied. "It is a niche market. Very lucrative." Bill looked at the house. The lights flickered. "What about the parents?" Bill asked. "They are inside eating pizza," Molly said. "They are blind to the outside." "And the security system?" "It is a motion sensor cat feeder," Molly said. "I hacked it. It thinks we are a large cat." "That is brilliant," Bill admitted. "So, we sneak in like burglars?" "We sneak in like guests," Molly corrected. "Wearing these." She pulled out two party hats. Bill stared at them. "I look ridiculous," Bill said. "You look festive," Molly said. "Festive thieves don't get caught." Bill put the hat on. It flopped over his eyes. "Okay. When do we go?" Bill asked. "When the clown starts juggling," Molly said. "It creates chaos." "Chaos is good," Bill said. "Chaos is profit," Molly said. "Stop eating the bears. We need the sugar." Bill tossed the bag. It hit the floor. "Sorry," Bill said. "I am nervous." "Don't be," Molly said. "The only thing harder than stealing gold is stealing a trampoline." "Is there a trampoline?" Bill asked. "There is now," Molly said. "Check the manifest." Bill checked tablet. He frowned. "Wow," Bill said. "That is heavy." "We have a van," Molly said. "It has a lift gate." "I am ready," Bill said. "Wait," Molly stopped him. "One more thing." "What?" Bill asked. "Don't touch the piƱata," Molly warned. "It is booby-trapped." "With what?" Bill asked. "Confetti," Molly said. "It is sticky. It ruins value." Bill nodded solemnly. He adjusted his hat. "Understood," Bill said. "No confetti." "Good," Molly said. "Let's move." They stepped out of the shadows. Music thumped from the house. Molly signaled Bill. He nodded. They walked toward the gate. "Wait," Bill whispered. "I see a dog." "Is it a guard dog?" Molly whispered. "No," Bill said. "It is a poodle in a tutu." "Ignore it," Molly said. "It is security." "Right," Bill said. "Tutu poodle." "Focus," Molly hissed. "The clown is juggling." "He is juggling knives," Bill said. "That is new," Molly said. "But it works." "I am moving," Bill said. "Go," Molly said. Bill ran. Molly followed. They reached the window. It was open. "Easy," Molly said. "Too easy," Bill said. "Trust me," Molly said. They climbed in. Room smelled of cake. Tree was huge. "Look," Bill whispered. "Grab the big ones," Molly said. "What about the small ones?" Bill asked. "Leave them," Molly said. "Too much weight." "Okay," Bill said. "Big ones only." Bill filled bag. Molly watched door. "How long?" Bill asked. "Five minutes," Molly said. "That is tight," Bill said. "We are tight," Molly said. Bill zipped bag. He stood up. "Done," Bill said. "Good," Molly said. "Let's go." They turned. Clown stood there. "Happy birthday," clown said. "We are leaving," Molly said. "No," clown said. "You are staying." "Bill," Molly said. "The grabber." "I got it," Bill said. He swung. Clown fell. "Nice shot," Molly said. "Thanks," Bill said. They ran. "Did we get it all?" Bill asked. "Mostly," Molly said. "Good enough," Bill said. "Always," Molly said.
created_at
"2026-02-25T22:17:10.432Z"
2/25/2026, 10:17:10 PM
model
"qwen/qwen3.5-27b-20260224"
app_id
182717
external_user
(null)
streamed
true
cancelled
false
latency
489
moderation_latency
(null)
generation_time
75761
tokens_prompt
77
tokens_completion
7495
native_tokens_prompt
76
native_tokens_completion
7534
native_tokens_completion_images
(null)
native_tokens_reasoning
6350
native_tokens_cached
0
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
0.0181044
router
(null)
provider_responses
0
endpoint_id
"f3f88554-0c22-4d57-973b-0dec948808e8"
id
"chatcmpl-00717d3f-faff-9dc6-9b31-fa1a011dd40a"
is_byok
false
latency
489
model_permaslug
"qwen/qwen3.5-27b-20260224"
provider_name
"Alibaba"
status
200
api_type
"completions"
id
"gen-1772057830-Q0L87O0pZyXUAmrxBSi5"
upstream_id
"chatcmpl-00717d3f-faff-9dc6-9b31-fa1a011dd40a"
total_cost
0.0181044
cache_discount
(null)
upstream_inference_cost
0
provider_name
"Alibaba"
Evaluation details
Result
Evaluator
Details
Meta Data
0.00%
Matches word count
n/a
neededClean
false
words
715
71.69%
Dialogue to Total Word Ratio
Ratio: 52.58%, Deviation: 2.58%
neededClean
false
wordsTotal
717
wordsDialogue
377
35.8432%