NC Bench
Overview
About
Tests
Categories
Models
Vendors
Compare
Run: 2026-07-09-01KX400PDJGJ26F7Y68HB68PFE
Run Details
Model:
GPT-5.6 Sol (Reasoning)
Test:
Dialogue tags
Scenario:
Write 500 words with 50% dialogue
Duration:
55.1s
Cost:
$0.151290
Provider:
OpenAI
Evaluation summary
Score
Evaluator
Details Link
100.00%
Matches word count
(details)
99.95%
Dialogue to Total Word Ratio
(details)
99.9753%
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 the birthday invitation across the café table, pinning each corner with stolen sugar packets while Bill suspiciously examined a balloon artist’s advertisement nearby. “The target is six years old. Security consists of twelve parents, one Labrador, and a magician whose trousers contain unexplained quantities of doves and handkerchiefs.” Bill stroked his moustache, which was false, upside down, and currently adhered with marmalade because he had misplaced the expensive criminal-grade spirit gum at breakfast. “We enter disguised as entertainers. I shall be Professor Pickle, balloon surgeon. You can be my glamorous assistant, Brenda the Ferociously Uninsured, for authenticity, Molly.” Molly stared until his moustache surrendered and dropped into his coffee. Bill retrieved it, squeezed it thoughtfully, and stuck it beneath his nose backward again. “No. Children detect fear, weakness, and budget entertainment instantly. We go as relatives. Every party includes one aunt nobody remembers inviting in the first place.” She produced two name badges: AUNT MOLLY and UNCLE WILLIAM. Bill flinched at his badge, as though respectability might spread through direct physical contact alone. “William is my tax-paying name. Besides, what gifts justify this peril? If it’s educational toys, I refuse completely, on both moral and professional grounds forever.” Molly unfolded a reconnaissance photograph showing a mountain of glittering packages. Atop them sat one enormous box wrapped in gold paper and guarded by cupcakes. “That contains Turbo Castle, this Christmas’s impossible toy. We steal everything else for camouflage, then sell the castle quietly abroad to desperate, wealthy grandparents later.” Bill leaned closer. His tie dipped into the coffee, absorbed half a cup, and slithered across the table like an exhausted, caffeinated garden snake escaping. “Counterproposal: we steal only wrapping paper. Nobody cries, the parents blame recycling, and we corner the festive stationery market before December arrives properly this year.” Molly rescued the photograph from his tie and outlined the floor plan. The presents waited in the conservatory, beside a pirate-themed cake and carrot sticks. “We cross the garden during musical statues. When everyone freezes, we keep moving very slowly. Humans cannot perceive criminals traveling at tortoise speed, probably, Bill.” Bill considered this, nodding with the solemnity of a judge. Outside, a waiter dropped three plates; Bill instantly crawled beneath the table and covered himself. “Excellent reflexes. Afterward, we load the presents into the catering van.” “Do either of us know how to drive legally, or just downhill very quickly?” Molly’s silence lasted long enough for the waiter to set another table. She amended the plan, replacing the van with a collapsible shopping trolley nearby. “We’ll need several trips. Unless the Labrador pulls it. I speak fluent dog.” “You once got bitten by a decorative cushion, Bill. It had tassels.” Bill emerged, wearing the tablecloth like a cape, and struck a heroic pose. The moustache slid downward again, finally attaching itself to his left eyebrow. “That cushion was hostile.” “Tomorrow, three o’clock. Bring the trolley, sensible shoes, and absolutely no marmalade whatsoever, William, unless you plan to attract wasps again.”
created_at
"2026-07-09T17:50:10.363Z"
7/9/2026, 5:50:10 PM
model
"openai/gpt-5.6-sol-20260709"
app_id
182717
external_user
(null)
streamed
true
cancelled
false
latency
3207
moderation_latency
(null)
generation_time
54430
tokens_prompt
77
tokens_completion
1544
native_tokens_prompt
66
native_tokens_completion
5032
native_tokens_completion_images
0
native_tokens_reasoning
4370
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
0.15129
router
(null)
provider_responses
0
endpoint_id
"a54c5de0-89bf-4ad7-a212-cf977eed918a"
id
"resp_0a290a91fbc88396016a4fdf5270008197880fce0273e50e33"
is_byok
false
latency
203
model_permaslug
"openai/gpt-5.6-sol-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-1783619410-IePc0P7nkfs4qVxu7FKZ"
session_id
(null)
preset_id
(null)
api_type
"completions"
id
"gen-1783619410-kie4Mh8C7aX7epCURFwe"
upstream_id
"resp_0a290a91fbc88396016a4fdf5270008197880fce0273e50e33"
total_cost
0.15129
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
500
99.95%
Dialogue to Total Word Ratio
Ratio: 49.90%, Deviation: 0.10%
neededClean
false
wordsTotal
503
wordsDialogue
251
99.9753%