NC Bench
Overview
About
Tests
Categories
Models
Vendors
Compare
Run: 2026-03-18-01KM07RB68C239C1DP93HWD0ZK
Run Details
Model:
GPT-5.4 Mini (Reasoning, Low)
Test:
Dialogue tags
Scenario:
Write 200 words with 10% dialogue
Duration:
4.0s
Cost:
$0.001868
Provider:
OpenAI
Evaluation summary
Score
Evaluator
Details Link
0.03%
Matches word count
(details)
99.61%
Dialogue to Total Word Ratio
(details)
49.8202%
Overall
Model input
User/Human
You are a author of comedic heist books. Please write a 200 word scene that has 10% 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 bakery van, studying the neon banner fluttering over the community hall: HAPPY BIRTHDAY, MAISIE! The sound of squealing children leaked through the doors like steam from an overworked kettle. Bill adjusted his cheap magician’s tuxedo and the absurdly tiny bow tie he had insisted was “respectable criminal camouflage.” “You look like a disappointed penguin,” Molly whispered. “I look,” Bill said, straightening his cuffs, “like a man with a plan.” Their plan involved cake, confusion, and the strategic misuse of a balloon animal sword. Molly had already memorized the gift table from the party host’s social media photos. Thirty-two presents, all labeled, stacked, and practically begging to be liberated. A scooter. Two remote-control dinosaurs. A suspiciously heavy box that Bill had decided was “probably electronics or remorse.” The clown at the entrance checked names with the grim devotion of a border guard. Molly smiled sweetly and carried a tray of cupcakes borrowed from a neighboring bake sale. Bill wheeled in a fake gift cart hidden beneath a tarp printed with cartoon stars. “Remember,” Molly murmured, “we blend in.” “With children?” “With panic,” she said. Inside, a chorus of shrieks rose as a magician failed to produce a rabbit and instead produced a fire extinguisher. Molly seized the moment, Bill bowed to the chaos, and together they glided toward the present mountain like professionals with very bad intentions.
created_at
"2026-03-18T10:27:11.19Z"
3/18/2026, 10:27:11 AM
model
"openai/gpt-5.4-mini-20260317"
app_id
182717
external_user
(null)
streamed
true
cancelled
false
latency
3389
moderation_latency
(null)
generation_time
3712
tokens_prompt
77
tokens_completion
510
native_tokens_prompt
66
native_tokens_completion
404
native_tokens_completion_images
0
native_tokens_reasoning
100
native_tokens_cached
0
num_media_prompt
(null)
num_input_audio_prompt
(null)
num_media_completion
0
num_search_results
(null)
origin
"https://ncbench.com/"
is_byok
false
finish_reason
"stop"
native_finish_reason
"completed"
usage
0.0018675
router
(null)
provider_responses
0
endpoint_id
"9ee065b2-3d1c-43bc-bdd7-28af3b148282"
id
"resp_07d599e820deb38c0169ba7dff3da8819790f7c475247aa3ea"
is_byok
false
latency
205
model_permaslug
"openai/gpt-5.4-mini-20260317"
provider_name
"OpenAI"
status
200
user_agent
"langchainjs-openai/1.0.0 ((node/v24.14.0; linux; x64))"
http_referer
(null)
api_type
"completions"
id
"gen-1773829631-OAG7yDuHCqQzG07Zho60"
upstream_id
"resp_07d599e820deb38c0169ba7dff3da8819790f7c475247aa3ea"
total_cost
0.0018675
cache_discount
(null)
upstream_inference_cost
0
provider_name
"OpenAI"
Evaluation details
Result
Evaluator
Details
Meta Data
0.03%
Matches word count
n/a
neededClean
false
words
230
99.61%
Dialogue to Total Word Ratio
Ratio: 12.50%, Deviation: 2.50%
neededClean
false
wordsTotal
232
wordsDialogue
29
49.8202%