NC Bench
Overview
About
Tests
Categories
Models
Vendors
Compare
Run: 2026-03-09-01KKACCZ42C1XPQ0MNSP8F4TE9
Run Details
Model:
GPT-5.4 (Reasoning, Low)
Test:
Dialogue tags
Scenario:
Write 200 words with 90% dialogue
Duration:
36.4s
Cost:
$0.038235
Provider:
OpenAI
Evaluation summary
Score
Evaluator
Details Link
100.00%
Matches word count
(details)
51.81%
Dialogue to Total Word Ratio
(details)
75.9059%
Overall
Model input
User/Human
You are a author of comedic heist books. Please write a 200 word scene that has 90% 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, this is low even for us.” “Bill, it’s not low; it’s efficient. Children can’t fence their presents.” “They can cry, though. Loudly.” “Which is why we arrive during cake. Frosting muffles outrage.” “What are we stealing?” “Everything with batteries, ribbons, or resale value.” “At a children’s party?” “Affluent parents panic-buy. Last year I saw a pony folded into a bag.” “You’re lying.” “I’m marketing the opportunity.” “What about disguises?” “Magicians. No one questions a top hat near balloons.” “I hate balloons. They squeak like guilty shoes.” “Then clown noses.” “I’d rather go to prison honestly.” “Fine. Caterers.” “Can you carry a tray without eating the evidence?” “Can you stop jingling when you’re nervous?” “That’s my lockpick set.” “Sounds like Santa developing a drinking problem.” “Timing?” “We enter at three, mingle by the bouncy castle, lift the loot during ‘Pass the Parcel,’ and vanish before the piñata.” “What if there’s a dog?” “We bribe it with sausage rolls.” “What if there are twins?” “Why would twins matter?” “Twins notice things. It’s their sport.” Molly sighed. “Bill, focus.” “I’m focused. Focused on ‘crime adjacent to a face painter.’” “Think bigger.” “Than stealing from children?” “Bill, we’re stealing from parents with receipts.”
created_at
"2026-03-09T22:45:03.756Z"
3/9/2026, 10:45:03 PM
model
"openai/gpt-5.4-20260305"
app_id
182717
external_user
(null)
streamed
true
cancelled
false
latency
35351
moderation_latency
(null)
generation_time
36181
tokens_prompt
77
tokens_completion
726
native_tokens_prompt
66
native_tokens_completion
2538
native_tokens_completion_images
0
native_tokens_reasoning
2227
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
0.038235
router
(null)
provider_responses
0
endpoint_id
"9ff5625c-403f-4d7f-b895-58ac7295062c"
id
"resp_0c5975ed84d3f5110169af4d6fd7988197a5cca93aa41657c1"
is_byok
false
latency
211
model_permaslug
"openai/gpt-5.4-20260305"
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-1773096303-RqTAjXH84uqhlVScf8In"
upstream_id
"resp_0c5975ed84d3f5110169af4d6fd7988197a5cca93aa41657c1"
total_cost
0.038235
cache_discount
(null)
upstream_inference_cost
0
provider_name
"OpenAI"
Evaluation details
Result
Evaluator
Details
Meta Data
100.00%
Matches word count
n/a
neededClean
false
words
200
51.81%
Dialogue to Total Word Ratio
Ratio: 99.00%, Deviation: 9.00%
neededClean
false
wordsTotal
201
wordsDialogue
199
75.9059%