NC Bench
Overview
About
Tests
Categories
Models
Vendors
Compare
Run: 2026-03-18-01KM0AHZQATWE1NYA9Z858EWT3
Run Details
Model:
GPT-5.4 Mini
Test:
Dialogue tags
Scenario:
Write 500 words with 30% dialogue
Duration:
6.5s
Cost:
$0.004946
Provider:
OpenAI
Evaluation summary
Score
Evaluator
Details Link
0.00%
Matches word count
(details)
56.60%
Dialogue to Total Word Ratio
(details)
28.3011%
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 pressed her face to the rain-smeared window of the van and watched the brightly lit community hall across the street. Paper balloons swayed from the doorway. A painted banner read HAPPY BIRTHDAY, CAPTAIN COSMO! in letters so cheerful they looked suspiciously judgmental. Bill sat in the driver’s seat, wearing gloves, a wool cap, and an expression of deep professional disappointment. “I still think this is beneath us,” he said. “It’s a job,” Molly said. She pulled a folded map from her coat and flattened it against her knee. “A profitable job.” “Stealing from a bank is a job. Swapping a duke’s diamonds is a job. This is theft from toddlers.” “From toddlers with presents,” Molly said. “Focus on the presents.” Bill glanced at the hall. Through the windows they could see a jungle of streamers, a magician in a gold vest pretending to be a space pirate, and at least thirty children rotating around a cake as though orbiting a very loud sun. “I’m focused. I’m just also haunted.” Molly grinned. “You’ll survive. The parents leave bags by the stage, we walk in with the caterer’s trolley, wheel out the gifts, and vanish before anyone notices little Captain Cosmo has been robbed blind.” Bill frowned. “You make it sound elegant. It sounds like two middle-aged raccoons with logistics.” Molly tapped the map. “Because it is elegant. We have timing, disguises, and a getaway route. We are artists.” “We’re criminals.” “Same thing if you do it cleanly.” Bill exhaled, then reached into the back and produced two bright paper crowns. One had glitter on it. “If I’m wearing a disguise, I want my dignity back afterward.” Molly took the crowns and inspected them. “These came with the van?” “I liberated them from a party supply store,” Bill said. “For morale.” “For morale,” Molly repeated, and placed the glitter crown on her head. It sat there with appalling cheer. “Fine. You can be King Bill of the Caterers.” “I hate this.” “You love this.” He did not answer, which in his case was the same as confessing. They waited until a burst of music and shrieking laughter came through the open side door of the hall. A clown stepped outside to smoke, then immediately noticed the van and stared at it with the solemn intelligence of a man who had learned too much too young. Molly slid lower in her seat. Bill murmured, “We’ve been judged by a clown.” “We’ll be judged by worse before sunset.” The clown flicked ash into a flowerpot and wandered off. Molly opened the rear doors and pulled the trolley into the alley. It was wrapped in black fabric with a neat row of silver plates stacked on top, a cover story so convincing it made the trolley look nervous. “Remember,” she said, “smile like you belong.” Bill lifted the trolley handles. “I never smile like I belong anywhere.” “Then try grimly helpful.” “That’s my natural face.” They moved toward the hall entrance as the caterer’s assistants bustled in and out carrying trays of juice boxes and miniature sandwiches. Molly gave a nod so brisk and competent that one of them stepped aside on instinct. Bill followed, shoulders hunched, his glitter crown tilted over one eyebrow. Inside, the party was a storm of sugar and noise. Children launched themselves at each other in capes and cardboard helmets. A bouncy castle thudded in the corner. Somewhere a parent was saying, “No, sweetie, that’s not your turn with the piñata, that’s the fire extinguisher.” Molly kept her voice low. “Stage left. Gift table is behind the moon cutout.” Bill steered the trolley through a knot of dancing children. “I see it.” “Count the bags.” “I’m counting the children who can identify us later.” “Bill.” “Fifteen gift bags, four wrapped boxes, one suspiciously large parcel.” “Excellent.” They reached the gift table. Molly began loading presents into the false compartments under the trolley’s silver plates with quick, practiced hands. “Birthday haul,” she whispered. “Easy money.” Bill snatched up a ribboned box and shook it. Something inside rattled like plastic dinosaurs. “One of these has a drum set.” “Then thank heaven we got here first.” A voice boomed over the music. “Who’s ready for cake?” Every child in the hall shrieked yes. The room surged toward the center table in a single ecstatic wave. Molly snapped the final compartment shut and smoothed the fabric. Bill stared at the trolley, now heavy with stolen gifts. “We are truly awful people.” Molly adjusted the glitter crown and nodded toward the door. “Yes,” she said, “but look how organized we are.” They rolled the trolley toward the exit just as the first slice of cake was cut, and the hall exploded in applause, frosting, and innocent joy. Outside, rain polished the pavement like a fresh getaway. Molly smiled, Bill groaned, and together they vanished into the night with the loot and absolutely no shame at all.
created_at
"2026-03-18T11:16:08.572Z"
3/18/2026, 11:16:08 AM
model
"openai/gpt-5.4-mini-20260317"
app_id
182717
external_user
(null)
streamed
true
cancelled
false
latency
428
moderation_latency
(null)
generation_time
5962
tokens_prompt
77
tokens_completion
1360
native_tokens_prompt
66
native_tokens_completion
1088
native_tokens_completion_images
0
native_tokens_reasoning
0
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
"completed"
usage
0.0049455
router
(null)
provider_responses
0
endpoint_id
"9ee065b2-3d1c-43bc-bdd7-28af3b148282"
id
"resp_0483e8e45ab5d8a80169ba8978a43c81948af630fa334eed63"
is_byok
false
latency
200
model_permaslug
"openai/gpt-5.4-mini-20260317"
provider_name
"OpenAI"
status
200
user_agent
"langchainjs-openai/1.0.0 ((node/v24.14.0; linux; x64))"
http_referer
(null)
api_type
"completions"
id
"gen-1773832568-owxQFlmQybHBciVnRVgF"
upstream_id
"resp_0483e8e45ab5d8a80169ba8978a43c81948af630fa334eed63"
total_cost
0.0049455
cache_discount
(null)
upstream_inference_cost
0
provider_name
"OpenAI"
Evaluation details
Result
Evaluator
Details
Meta Data
0.00%
Matches word count
n/a
neededClean
false
words
822
56.60%
Dialogue to Total Word Ratio
Ratio: 33.37%, Deviation: 3.37%
neededClean
false
wordsTotal
824
wordsDialogue
275
28.3011%