NC Bench
Overview
About
Tests
Categories
Models
Vendors
Compare
Run: 2026-03-11-01KKFG4VGX0J1HGE85JEPDJSE6
Run Details
Model:
Nemotron 3 Nano
Test:
Dialogue tags
Scenario:
Write 500 words with 70% dialogue
Duration:
30.1s
Cost:
$0.000305
Provider:
DeepInfra
Evaluation summary
Score
Evaluator
Details Link
0.00%
Matches word count
(details)
0.00%
Dialogue to Total Word Ratio
(details)
0.0014%
Overall
Model input
User/Human
You are a author of comedic heist books. Please write a 500 word scene that has 70% 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: Bill! Bill, look at this spreadsheet! I aced the guest list analytics! Bill: Molly, we agreed "glitter analytics" was a joke. Also, "cupcake calorie density" is not a real metric. Molly: It is now! See? Kids over seven get triple frosting. That's, like, a 200% ROI on stolen sprinkles. Bill: ROI? We're talking about a *birthday party*, Molly! Not a dragon's hoard. What if Mrs. Henderson's Aunt Carol brings her famous jell-o salad? Molly: Jell-o Salad? Bill, that's a *distraction*! We use it as camouflage! While they're busy saying "pass the jell-o," we're swapping the piƱata with a decoy filled with *squirrel pellets*. Bill: Squirrel pellets? That's not a decoy, that's a war crime! Molly: It's *psychological* warfare! They'll think it's a trap! Also, I saw a kid with a slingshot shaped like a taco. Terrifying. Bill: Okay, fine, taco-tac slingshots are concerning. But the *real* problem is the cake. It's a *three-tiered* cake. Mounted on a *pedestal*. Molly: I've got this! I practiced the "Triple Squeeze" maneuver! One squeeze for each tier! Like... like milking a cow but with cake! Bill: You practiced on a toy cow! And it's *marble*. You'll shatter it! Molly: Then we get the pieces! They'll never notice! Unless... Bill? Bill, did you just sneeze on the frosting sample? Bill: Yeah! Gross, right? But strategic! It masks the vanilla scent of our getaway van's interior. Smells like... birthday. Freedom. Molly: Bill, focus! The *real* plan involves the cake. You distract Karen with... uh... is it "dust storm simulations"? Bill: I was going to say "compliment her tote bag"? But Karen carries *seven* tote bags. It might take longer than reprogramming the Roomba! Molly: Too late! I bribed the Roomba with a gummy worm! It's now dancing to *Baby Shark* on loop! Gives us, like, twelve minutes! Bill: Twelve minutes? Molly, our getaway plan involves me carrying the cake *and* you carrying the loot bag *and* us hopping that pothole by the playground! Molly: We can *run*! And if we trip, we blame it on the imaginary squirrel! Squirrel's our scapegoat! Bill: ...Maximum squirrel credibility. Got it. But if the toddlers see us, Molly, *nothing* draconian. No egging the bounce house. No yelling "FIRE!" in the bubble wrap zone. Molly: Roger doodle! I'm just here to snag the *fun-sized* peanut butter cups *after* the main haul. It's like bonus points! Bill: This feels wrong. Stealing from *children*... Molly: Stealing *for* children! We're redistributing joy! I'll save a cupcake for the widdle weasels in pigtails! Bill: Ugh. Fine. But if we're caught, you take the blame. You've got that "incredibly earnest" face. Molly: Deal! Now, let's go! The slide's waiting! Bill: Molly... Molly: *Yes*, Bill? Bill: ...Did you just put peanut butter on the whiteboard? Molly: ...It was stressed. Volunteer your face for testing!
created_at
"2026-03-11T22:26:44.395Z"
3/11/2026, 10:26:44 PM
model
"nvidia/nemotron-3-nano-30b-a3b"
app_id
182717
external_user
(null)
streamed
true
cancelled
false
latency
451
moderation_latency
(null)
generation_time
29976
tokens_prompt
77
tokens_completion
1418
native_tokens_prompt
82
native_tokens_completion
1506
native_tokens_completion_images
(null)
native_tokens_reasoning
672
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
(null)
native_finish_reason
(null)
usage
0.0003053
router
(null)
provider_responses
0
endpoint_id
"2bd3912d-8b35-4818-b873-24ccfefa6cf9"
id
"chatcmpl-R7MFkr9xRF4joQNd083IZXl2"
is_byok
false
latency
54
model_permaslug
"nvidia/nemotron-3-nano-30b-a3b"
provider_name
"DeepInfra"
status
200
user_agent
"langchainjs-openai/1.0.0 ((node/v24.14.0; linux; x64))"
http_referer
(null)
api_type
"completions"
id
"gen-1773268004-iKEwkAb9Y3mXrF5CZm3V"
upstream_id
"chatcmpl-R7MFkr9xRF4joQNd083IZXl2"
total_cost
0.0003053
cache_discount
(null)
upstream_inference_cost
0
provider_name
"DeepInfra"
Evaluation details
Result
Evaluator
Details
Meta Data
0.00%
Matches word count
n/a
neededClean
false
words
468
0.00%
Dialogue to Total Word Ratio
Ratio: 4.45%, Deviation: 65.55%
neededClean
false
wordsTotal
472
wordsDialogue
21
0.0014%