NC Bench
Overview
About
Tests
Categories
Models
Vendors
Compare
Run: 2026-03-18-01KM08MWBYT2CEH8Q828B9ZBHZ
Run Details
Model:
GPT-5.4 Nano
Test:
Dialogue tags
Scenario:
Write 200 words with 90% dialogue
Duration:
2.2s
Cost:
$0.000521
Provider:
OpenAI
Evaluation summary
Score
Evaluator
Details Link
0.00%
Matches word count
(details)
36.79%
Dialogue to Total Word Ratio
(details)
18.3940%
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, check the frosting map again. If we hit the wrong house, we’re paying in sprinkles.” “Relax, Bill. My frosting map is flawless. The driveway is shaped like a chocolate smile.” “Yeah, and last time you called a decoy balloon ‘a chocolate smile’ and it was actually a helium octopus.” “Don’t blame me. Octopuses are naturally mischievous.” “We’re stealing presents from a children’s party, not auditioning for a puppet show.” “Correction: we’re stealing presents that are already basically ours in spirit. Those socks are doomed.” “Spirited socks?” “Shh. Listen. Kids are cheering. That means the parents are distracted by cake.” “I hate when cake becomes a weapon. What’s our entry plan?” “We slide under the porch, pop the gift-guest list lock, and—” “Stop saying ‘lock’ like you have a tool bag full of magic.” “I do. It’s called confidence.” “Confidence doesn’t fit through vents.” “Then we don’t use vents. We use the snack corridor. Follow me.” “Snack corridor sounds like a kidnapping documentary.” “Better than ‘failed heist,’ which is your favorite sequel.” “Fine. What do we steal first?” “Everything wrapped in shiny paper. Also, that dinosaur remote-control thing.” “Dinosaur remote-control is overkill.” “Bill, it’s for stealth. It distracts them with chomps.” “Last question: how do we leave without setting off the doorbell?” “We don’t. We ring it after we’re gone.” “After? That’s… petty.” “Exactly. Now quiet—act innocent.” “Which means what, exactly?” “Smile like a birthday clown.” “I don’t have clown hands.” “Use your thief hands. They’re already creepy.”
created_at
"2026-03-18T10:42:46.281Z"
3/18/2026, 10:42:46 AM
model
"openai/gpt-5.4-nano-20260317"
app_id
182717
external_user
(null)
streamed
true
cancelled
false
latency
443
moderation_latency
(null)
generation_time
1982
tokens_prompt
77
tokens_completion
490
native_tokens_prompt
66
native_tokens_completion
406
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
0.0005207
router
(null)
provider_responses
0
endpoint_id
"0c835f2e-c18d-4e8c-b245-e1e3bd08b97f"
id
"resp_0e29d91f5e9f5ba40169ba81a656fc81948e8b6d46123ca540"
is_byok
false
latency
191
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-1773830566-USUg5IVvxSefSo3Q3ycN"
upstream_id
"resp_0e29d91f5e9f5ba40169ba81a656fc81948e8b6d46123ca540"
total_cost
0.0005207
cache_discount
(null)
upstream_inference_cost
0
provider_name
"OpenAI"
Evaluation details
Result
Evaluator
Details
Meta Data
0.00%
Matches word count
n/a
neededClean
false
words
249
36.79%
Dialogue to Total Word Ratio
Ratio: 100.00%, Deviation: 10.00%
neededClean
false
wordsTotal
252
wordsDialogue
252
18.3940%