NC Bench
Overview
About
Tests
Categories
Models
Vendors
Compare
Run: 2026-07-09-01KX40SRZ2NWFDMFQYEJ2WF5S6
Run Details
Model:
GPT-5.6 Terra (Reasoning)
Test:
Dialogue tags
Scenario:
Write 500 words with 50% dialogue
Duration:
24.5s
Cost:
$0.046365
Provider:
OpenAI
Evaluation summary
Score
Evaluator
Details Link
100.00%
Matches word count
(details)
89.48%
Dialogue to Total Word Ratio
(details)
94.7392%
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 crouched behind the community hall’s recycling bins, studying balloons bobbing above the fence like cheerful surveillance devices designed by an exceptionally optimistic police department. “Bill, if you squeak that rubber duck again, I will personally donate you to the toddler raffle, and nobody will bid above seventy pence either.” Bill held the duck beneath his trench coat, where it emitted a muffled quack whenever he breathed, sabotaging his reputation as a professional burglar entirely. “I cannot help breathing, Molly. It is traditionally considered useful. Besides, the duck is our distraction, our cover story, and possibly our getaway vehicle tonight.” Beyond the fence, children shrieked around a magician whose cape repeatedly caught on folding chairs, while a mountain of wrapped presents glittered beside the cake. “Forget the duck. We enter as entertainers, smile at the parents, locate the gift table, and leave before anyone notices their offspring have acquired joy.” Molly unfolded Bill’s crayon-colored blueprint, drawn during breakfast, featuring arrows, a pirate ship, several questionable trapdoors, and a note reading, “beware sticky bits in there.” “Your map has a dragon guarding the bouncy castle, Bill. Is that intelligence, imagination, or lingering trauma from my niece’s sixth birthday party last year?” A gust lifted the blueprint into a puddle, where the dragon dissolved, leaving only the route through the kitchen and alarming sketch of a sausage. “Kitchen route is excellent. Parents congregate there, discussing mortgages and coffee machines, which means nobody guards presents except perhaps the birthday child and her corgi.” Bill peered through a gap in the fence. The corgi wore a paper crown and inspected each guest with the grim concentration of customs officials. “Corgis are low to the ground, Molly. Their authority is largely horizontal. I can distract him with cheese, compliments, or a tiny forged passport from Wales.” Molly adjusted her sequined waistcoat, stolen from a karaoke host, and produced two balloons twisted into swords, their handles already slick with mysterious frosting residue. “We need a better disguise than matching waistcoats and balloon swords. You look like a pirate who lost custody of his parrot in court yesterday.” Inside, the magician accidentally pulled a real rabbit from his sleeve, then apologised to it, while adults applauded with the exhausted hope of rescue soon. “Fine. We pose as balloon security. Every party has balloon security now. It is the fastest-growing field in childhood event management, after emergency glitter removal.” Bill nodded solemnly, as though this nonsense were printed in a handbook, then strapped the rubber duck to his belt like an official identification badge. “When we reach the presents, take only the big ones. Small boxes are socks, puzzles, or worse, educational gifts. We have standards, however flexible today.” Molly glanced at the gift mountain, where one enormous package trembled beneath dinosaur paper. A label announced it was for Beatrice, age seven, “from Nana.” “Actually, Bill, we steal nothing. We replace every tag, make the magician look guilty, and demand payment for solving the crime. Bring the glitter cannon.”
created_at
"2026-07-09T18:03:52.169Z"
7/9/2026, 6:03:52 PM
model
"openai/gpt-5.6-terra-20260709"
app_id
182717
external_user
(null)
streamed
true
cancelled
false
latency
3014
moderation_latency
(null)
generation_time
24207
tokens_prompt
77
tokens_completion
1364
native_tokens_prompt
66
native_tokens_completion
3080
native_tokens_completion_images
0
native_tokens_reasoning
2435
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
0.046365
router
(null)
provider_responses
0
endpoint_id
"3cc89000-ae34-4dde-9c7d-5ce009c8af0b"
id
"resp_0bd45167355997d7016a4fe28841b081938bc5077484c7351c"
is_byok
false
latency
281
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-1783620232-qYwDoxHb56v3w8IFPWxT"
session_id
(null)
preset_id
(null)
api_type
"completions"
id
"gen-1783620232-V3LCinRpx5aonOHxg4id"
upstream_id
"resp_0bd45167355997d7016a4fe28841b081938bc5077484c7351c"
total_cost
0.046365
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
100.00%
Matches word count
n/a
neededClean
false
words
501
89.48%
Dialogue to Total Word Ratio
Ratio: 51.49%, Deviation: 1.49%
neededClean
false
wordsTotal
503
wordsDialogue
259
94.7392%