NC Bench
Overview
About
Tests
Categories
Models
Vendors
Compare
Run: 2026-03-18-01KM07RR0Y87SZK9V862PN4R3X
Run Details
Model:
GPT-5.4 Nano
Test:
Dialogue tags
Scenario:
Write 200 words with 90% dialogue
Duration:
2.6s
Cost:
$0.000443
Provider:
OpenAI
Evaluation summary
Score
Evaluator
Details Link
0.21%
Matches word count
(details)
50.02%
Dialogue to Total Word Ratio
(details)
25.1195%
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, are you sure this tiny door is for the party house, or did we pick the wrong address again?” Bill whispered. “It’s the right house. The big inflatable dinosaur is basically a beacon. Now shut up and listen—do you hear carols?” “Yeah, and I think the carols are coming from the living room with the glass chandelier. Who puts a chandelier in a kids’ party?” “People with money and questionable taste. We slip in, we grab presents, we vanish. Like professionals.” “Professionals? Molly, you called your lockpick ‘Pickles’ yesterday.” “That was an affectionate nickname. Okay, strategy: you distract the host.” “I don’t do distractions. I do sneaky breathing and suspiciously empty hands.” “Perfect. I’ll do the stealing.” “No, you’ll do the stealing and I’ll do the escaping?” “Don’t negotiate. Kids are distracted by balloons, adults by cake, and I’m both sweet and unstoppable.” “You’re not unstoppable. You got stuck in the hedge last time.” “I was practicing stealth shrubbery.” “Fine. But what if the presents are… in front of the children?” “Then we take the biggest ones first. Children love big. It’s science.” “Is it science or bribery with stolen toys?” “Both. Now, on my cue: you open the window.” “I can’t open the window. It’s locked.” “Then pick it.” “I can’t pick it. It has a baby sticker.” “Bill. Please. We’re stealing. Stop respecting stickers.”
created_at
"2026-03-18T10:27:24.332Z"
3/18/2026, 10:27:24 AM
model
"openai/gpt-5.4-nano-20260317"
app_id
182717
external_user
(null)
streamed
true
cancelled
false
latency
1096
moderation_latency
(null)
generation_time
2401
tokens_prompt
77
tokens_completion
410
native_tokens_prompt
66
native_tokens_completion
344
native_tokens_completion_images
0
native_tokens_reasoning
0
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
0.0004432
router
(null)
provider_responses
0
endpoint_id
"0c835f2e-c18d-4e8c-b245-e1e3bd08b97f"
id
"resp_064e42ae878d73d30169ba7e0c65ac819794e23c55bdcf8b77"
is_byok
false
latency
199
model_permaslug
"openai/gpt-5.4-nano-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-1773829644-geu9ZEbvGEf29BVsZRTr"
upstream_id
"resp_064e42ae878d73d30169ba7e0c65ac819794e23c55bdcf8b77"
total_cost
0.0004432
cache_discount
(null)
upstream_inference_cost
0
provider_name
"OpenAI"
Evaluation details
Result
Evaluator
Details
Meta Data
0.21%
Matches word count
n/a
neededClean
false
words
228
50.02%
Dialogue to Total Word Ratio
Ratio: 99.12%, Deviation: 9.12%
neededClean
false
wordsTotal
228
wordsDialogue
226
25.1195%