NC Bench
Overview
About
Tests
Categories
Models
Vendors
Compare
Run: 2026-09-23-01M37HW3DZ6VW7T473AK75P6H9
Run Details
Model:
GPT-6 Sol (Reasoning, Medium)
Test:
Dialogue tags
Scenario:
Write 500 words with 50% dialogue
Duration:
18.4s
Cost:
$0.009762
Provider:
OpenAI
Evaluation summary
Score
Evaluator
Details Link
1.04%
Matches word count
(details)
14.93%
Dialogue to Total Word Ratio
(details)
7.9846%
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 spread a hand-drawn floor plan across the café table. It was decorated with crayon suns, because she had borrowed both the paper and the crayons from her nephew. “The presents will be beside the cake,” she said. “We go in, collect them all, and leave before anyone notices.” Bill studied the drawing. “Why is the kitchen marked HERE BE DRAGONS?” “My nephew says the cook has a beard.” Bill drew a nervous circle around the kitchen. Their target was a seventh birthday party at a community hall. From the invitation, Molly had learned the theme was pirates. “We’ll wear costumes,” she said. “Nobody questions pirates carrying treasure.” “They might question pirates in their forties.” “Then act like you’re somebody’s uncle.” “I am somebody’s uncle. It hasn’t made me less suspicious.” Molly pulled two striped shirts from her bag. Bill held one against himself and grimaced. The sleeves ended well above his wrists; it had once belonged to her nephew. “Do you have a larger one?” he asked. “No. But I brought you an eye patch.” “What does that solve?” “People won’t look at your sleeves.” Bill put on the patch. For a moment he looked menacing, until he tried to reach for his coffee and knocked over the sugar bowl. “Depth perception,” Molly said. “We’ll work on it.” A waitress passed their table, glanced at the map, and kept walking. Molly lowered her voice anyway. She had never planned a theft in public before. “How are we getting invited?” Bill asked. “Your nephew going?” “No. I’ve been hired to entertain the children.” Bill stared. “Since when do you entertain children?” “Since the magician cancelled this morning.” “What are you going to do instead?” Molly produced a deck of cards. Three cards slid out of the box and into Bill’s coffee. She considered the brown, swelling corners of them. “Mostly,” she said, “I’m going to keep everyone looking at me.” Bill fished out a card. It was the seven of clubs. He laid it on a napkin as if preserving evidence of a terrible accident. “And while they look at you, I take the presents?” “Yes. There’s a trolley behind the hall.” “Won’t somebody notice a man wheeling away every gift?” “You’ll be a pirate. Call it a raid.” He pictured himself declaring a raid to twenty children, all of whom would know precisely how many presents they had brought. Molly, meanwhile, was practising a card flourish beneath the table. “We could take the cake instead,” Bill said. “We agreed. All the presents.” “I know. I just thought the cake might fit on the trolley.” Molly paused. She turned the invitation over and found the line she had missed: Parents welcome to stay. Bill watched her read it twice. “How many parents?” he asked. “Probably a few.” “You mean every child comes with a witness?” Molly folded the floor plan and tucked it away. Then she reached into her bag for a balloon pump. “We’ll need a better distraction,” she said. Bill removed the eye patch. “Please tell me you can make balloon swords.” “I can make one shape.” “What shape?” Molly began to pump. “Long.”
created_at
"2026-09-23T16:34:08.455Z"
9/23/2026, 4:34:08 PM
model
"openai/gpt-6-sol-20260922"
app_id
182717
external_user
(null)
streamed
true
cancelled
false
latency
6416
moderation_latency
(null)
generation_time
18370
tokens_prompt
75
tokens_completion
965
native_tokens_prompt
66
native_tokens_completion
963
native_tokens_completion_images
0
native_tokens_reasoning
285
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
0.009762
router
(null)
user_agent
"langchainjs-openai/1.0.0 ((node/v24.20.0; linux; x64))"
http_referer
(null)
request_id
"req-1790181248-k4zAdj26N1mkd22cF9zA"
session_id
(null)
preset_id
(null)
api_type
"completions"
id
"gen-1790181248-q298rvatv2FporDZHxHk"
upstream_id
"resp_0327417fb366b92d016ab3ff80935c87d192a95ea40d8b1485"
provider_responses
0
endpoint_id
"1cfc7d9d-4404-4b8e-9ee4-58ae45c9dcd4"
id
"resp_0327417fb366b92d016ab3ff80935c87d192a95ea40d8b1485"
is_byok
false
latency
511
model_permaslug
"openai/gpt-6-sol-20260922"
provider_name
"OpenAI"
status
200
total_cost
0.009762
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
1.04%
Matches word count
n/a
neededClean
false
words
526
14.93%
Dialogue to Total Word Ratio
Ratio: 43.83%, Deviation: 6.17%
neededClean
false
wordsTotal
527
wordsDialogue
231
7.9846%