NC Bench
Overview
About
Tests
Categories
Models
Vendors
Compare
Run: 2026-09-23-01M37HPY46858NKXYPXZA01BFC
Run Details
Model:
GPT-6 Sol
Test:
Dialogue tags
Scenario:
Write 500 words with 30% dialogue
Duration:
14.9s
Cost:
$0.006812
Provider:
OpenAI
Evaluation summary
Score
Evaluator
Details Link
100.00%
Matches word count
(details)
7.01%
Dialogue to Total Word Ratio
(details)
53.5046%
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 a hand-drawn map across the café table. It showed the Willow Street community hall, three exits, one suspiciously large cake, and a square labeled PRESENT MOUNTAIN. Crayon arrows crossed the page in every direction. Bill squinted at it. “Why is the cake marked dangerous?” “Because it’s a castle cake. Structural collapse is possible.” “Have you ever seen a cake collapse?” “Not twice.” Outside, rain tapped the windows. Inside, the lunch crowd ignored the two adults studying a children’s birthday party invitation with the concentration of generals planning a siege. The invitation was printed on glittery cardstock. Molly had found it on her doormat that morning. She lived across the hall from the birthday child’s grandmother, who often confused their mailboxes and, once, their casserole dishes. Molly had meant to return it. Then she noticed the party was that afternoon, and the idea had struck her as irresistibly silly. “The haul goes on this table,” she said, tapping the map. “We walk in, take every present, and leave.” “In broad daylight?” “Unless you know how to make a children’s party happen at midnight.” Bill considered the plan. “Costumes?” “Of course. I’m going as a magician. You’re my assistant.” “Why am I the assistant?” Molly pointed to the red sequined bow tie protruding from Bill’s coat pocket. “That proves nothing,” he said. At two o’clock, they reached the hall. Bill wore a cape borrowed from Molly’s nephew, fastened with a safety pin. Molly carried a top hat full of plastic flowers and a wand that lit up when she pressed a button. The hall doors swung open before either thief knocked. A woman wearing a paper crown looked them over. Behind her, children raced between tables while a weary man tried to stop a balloon arch from drifting into the ceiling fan. “Entertainment!” the woman cried. “You’re early. Wonderful.” Molly bowed. Bill attempted the same and caught his cape under one shoe. They were swept inside before they could explain. The presents stood exactly where Molly’s map promised: a glorious, teetering pile beside the cake. Unfortunately, twenty children stood between them and it. Molly cleared her throat. “Behold! A magic trick!” Every child turned. Bill whispered, “Which trick?” “The one where you help me.” “I hate that one.” Molly waved her wand over the hat. It flashed red, then blue. She reached inside and produced a plastic daisy. A small boy applauded politely. “Again,” he demanded, “but with a dinosaur.” For the next hour, Molly produced flowers, coins, and, accidentally, Bill’s bus pass. Bill made balloon animals so abstract that one girl named hers “Concern.” The children loved them. Molly laughed hard enough to forget about the presents. At the end, the birthday girl offered them each a cupcake. The woman in the crown thanked them twice and asked for their card. Outside, Bill unfastened his cape. “So. We stole nothing.” Molly looked at the cupcake in her hand. “Technically, these were given to us.” “Worst heist ever.” “Best cake, though.”
created_at
"2026-09-23T16:31:19.183Z"
9/23/2026, 4:31:19 PM
model
"openai/gpt-6-sol-20260922"
app_id
182717
external_user
(null)
streamed
true
cancelled
false
latency
3756
moderation_latency
(null)
generation_time
14813
tokens_prompt
75
tokens_completion
830
native_tokens_prompt
66
native_tokens_completion
668
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
0.006812
router
(null)
user_agent
"langchainjs-openai/1.0.0 ((node/v24.20.0; linux; x64))"
http_referer
(null)
request_id
"req-1790181079-VNNjD0nERkpY2cQ5MNf7"
session_id
(null)
preset_id
(null)
api_type
"completions"
id
"gen-1790181079-aJlTxXotJqVxEcDZiP3M"
upstream_id
"resp_0cf2b680c3050ad7016ab3fed751b887d19dd31c69c8b7e9e2"
provider_responses
0
endpoint_id
"1cfc7d9d-4404-4b8e-9ee4-58ae45c9dcd4"
id
"resp_0cf2b680c3050ad7016ab3fed751b887d19dd31c69c8b7e9e2"
is_byok
false
latency
2399
model_permaslug
"openai/gpt-6-sol-20260922"
provider_name
"OpenAI"
status
200
total_cost
0.006812
cache_discount
(null)
upstream_inference_cost
0
provider_name
"OpenAI"
response_cache_source_id
(null)
data_region
"global"
workspace_id
"97e315e5-d303-487d-83c1-83180e8a13d4"
Evaluation details
Result
Evaluator
Details
Meta Data
100.00%
Matches word count
n/a
neededClean
false
words
501
7.01%
Dialogue to Total Word Ratio
Ratio: 22.71%, Deviation: 7.29%
neededClean
false
wordsTotal
502
wordsDialogue
114
53.5046%