NC Bench
Overview
About
Tests
Categories
Models
Vendors
Compare
Run: 2026-07-09-01KX3ZZE0TS7AP96RHZVM5HE7K
Run Details
Model:
GPT-5.6 Terra
Test:
Dialogue tags
Scenario:
Write 500 words with 70% dialogue
Duration:
14.4s
Cost:
$0.016080
Provider:
OpenAI
Evaluation summary
Score
Evaluator
Details Link
0.00%
Matches word count
(details)
99.68%
Dialogue to Total Word Ratio
(details)
49.8385%
Overall
Model input
User/Human
You are a author of comedic heist books. Please write a 500 word scene that has 70% 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
“Children’s party,” Molly said, peering through binoculars from behind the hedge. “Ten o’clock. Garden venue. Bouncy castle. Clown. Cake shaped like a pirate ship. Estimated presents: forty-seven.” Bill adjusted his black ski mask. “You said this was a jewel job.” “It is. Plastic jewels, probably. But children have rich grandparents. There could be scooters. Tablets. Gift cards. One of those miniature electric cars.” “I don’t like it.” “You robbed a museum last month.” “Museums don’t scream when you take their property.” “Children scream if you look at them wrong. We’ll be gone before the candle ceremony.” Bill looked over the hedge. “That’s a lot of parents.” “Parents are distracted by free prosecco and competitive small talk.” “What’s the entry plan?” “You’re the clown.” Bill stared at her. “No.” “You’ve got the build.” “I’ve got the build of a man who has made regrettable decisions. That is different.” “The hired clown is late. I intercepted his van and told him there was a gas leak at the venue.” “You intercepted a clown?” “I flashed a badge.” “You don’t have a badge.” “I have three badges. Parking enforcement, animal control, and something in Latvian.” Bill sighed. “And you?” “Magician’s assistant.” “You don’t look like a magician’s assistant.” “I look mysterious.” “You look like you’re attending a funeral for a vampire.” “Black sequins are traditional.” A shriek rose from the garden, followed by the sound of a whistle. Bill winced. “What if they ask me to make balloon animals?” “Make a snake.” “What if they ask for a dog?” “Make a longer snake with legs.” “What if they ask why the dog has fangs?” “Tell them it’s a tax dog.” “That means nothing.” “Children adore nonsense. That’s why they watch magicians.” Bill pulled a red wig from Molly’s bag. “This smells like wet carpet and betrayal.” “It’s authentic.” “Whose clown nose is this?” “Don’t ask questions you can’t afford answers to.” He put on the wig. “How do I look?” “Like a man who owes a circus money.” “Excellent. Criminally persuasive.” Molly checked her watch. “Right. You enter through the side gate, announce a magic show, create chaos. I slip into the conservatory, locate the gift table, load the sack, and exit through the kitchen.” “What sort of chaos?” “Harmless chaos.” “You said that before the museum job.” “And no one was harmed.” “The curator fell into a sarcophagus.” “He emerged spiritually improved.” Bill glanced at the party. “There’s a pony.” “Unexpected variable.” “There’s always a pony when something goes wrong.” “Don’t engage with the pony.” “What if it engages with me?” “Offer it a carrot.” “I don’t have a carrot.” “Then don’t look edible.” Bill took a breath. “Fine. But if a child asks me to juggle, I’m leaving.” “You can juggle.” “I can juggle stolen watches in a quiet room.” “Same principle. Three objects, moving hands, false confidence.” They slipped through the gate. A woman in a floral dress spotted Bill and clapped. “Oh, marvelous! The clown’s here!” Bill gave a stiff wave. “Yes. Hello, tiny future taxpayers.” Molly hissed, “No taxes.” “Right. Hello, tiny... people.” Children swarmed him immediately. “Make a balloon dinosaur!” “Do a trick!” “Why is your face sad?” Bill looked at Molly over the sea of small hands. “Proceed,” he muttered. Molly glided toward the conservatory, passing a table piled high with wrapped boxes. There were dinosaurs, rockets, glittery bags, and one enormous parcel shaped suspiciously like a washing machine. Bill’s voice boomed behind her. “For my first trick, I shall make this handkerchief disappear!” A child shouted, “That’s your sock!” “Exactly. Observe closely.” Molly filled the sack quickly. A handheld game console. Two scooters. Three envelopes. A stuffed unicorn the size of a Labrador. Then a small girl appeared in the doorway. “Are you stealing Leo’s presents?” Molly froze. “No,” she said smoothly. “I’m testing them.” “For what?” “Security.” The girl narrowed her eyes. “That’s what my dad says when he eats my crisps.” Molly reached into her pocket and produced a silver coin. “Would you like to see a magic trick?” The girl nodded. Molly palmed the coin, flicked her wrist, and produced it from behind the girl’s ear. The girl gasped. “Do it again!” “Certainly. But first, perhaps you could distract the clown.” “He’s bad at balloons.” “Tell him you want a tax dog.” The girl ran off. From outside, Bill cried, “Molly! What is a tax dog?” Molly hoisted the sack and smiled. “A very expensive snake with legs.”
created_at
"2026-07-09T17:49:28.993Z"
7/9/2026, 5:49:28 PM
model
"openai/gpt-5.6-terra-20260709"
app_id
182717
external_user
(null)
streamed
true
cancelled
false
latency
613
moderation_latency
(null)
generation_time
14300
tokens_prompt
77
tokens_completion
1437
native_tokens_prompt
66
native_tokens_completion
1061
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)
num_fetches
(null)
web_search_engine
(null)
origin
"https://ncbench.com/"
is_byok
false
finish_reason
"stop"
native_finish_reason
"completed"
service_tier
"default"
usage
0.01608
router
(null)
provider_responses
0
endpoint_id
"3cc89000-ae34-4dde-9c7d-5ce009c8af0b"
id
"resp_0e0b5cc4b0e50592016a4fdf29121c8190891f41855ea72468"
is_byok
false
latency
232
model_permaslug
"openai/gpt-5.6-terra-20260709"
provider_name
"OpenAI"
status
200
user_agent
"langchainjs-openai/1.0.0 ((node/v24.18.0; linux; x64))"
http_referer
(null)
request_id
"req-1783619368-m49aqndnU4aYZ964TnTw"
session_id
(null)
preset_id
(null)
api_type
"completions"
id
"gen-1783619368-NrmxWLq7XsrYFIOBIN4b"
upstream_id
"resp_0e0b5cc4b0e50592016a4fdf29121c8190891f41855ea72468"
total_cost
0.01608
cache_discount
(null)
upstream_inference_cost
0
provider_name
"OpenAI"
response_cache_source_id
(null)
data_region
"global"
Evaluation details
Result
Evaluator
Details
Meta Data
0.00%
Matches word count
n/a
neededClean
false
words
746
99.68%
Dialogue to Total Word Ratio
Ratio: 69.75%, Deviation: 0.25%
neededClean
false
wordsTotal
747
wordsDialogue
521
49.8385%