NC Bench
Overview
About
Tests
Models
Vendors
Run: 2026-02-16-01KHM62WHYY6E2KJ0CWDW2X3CB
Run Details
Model:
DeepSeek V3.2
Test:
Dialogue tags
Scenario:
Write 500 words with 50% dialogue
Duration:
52.2s
Cost:
$0.000333
Provider:
Phala
Model input
User/Human
You are a author of comedic heist books. Please write a 500 word scene that has 50% 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
The air in the derelict warehouse office was thick with the smell of stale coffee and poor life choices. Molly, her face illuminated by the blue glow of a tablet, traced a finger over a schematized backyard. Bill meticulously sharpened a throwing knife with a file, the *scritch-scritch* sound the only punctuation to her briefing. “Okay, final run-through,” Molly said, not looking up. “The target is a sixth birthday party for one ‘Lucas Henderson,’ son of tech mogul Richard Henderson. Saturday, 2 PM, at their Crestwood estate. Perimeter is a six-foot fence, but it’s decorative. The real issue is the guest list: twenty-four six-year-olds, hyped on sugar and superheroes.” Bill nodded, testing the knife’s edge against his thumb. “Loud. Unpredictable. Perfect cover. What’s the exfil?” “Chaos,” Molly said, a ghost of a smile on her lips. “Pinata time is at 3:30. We go in as ‘Sparkle the Clown’ and her assistant ‘Bungles.’ You’ll be handling the donkey-tail-pinning game. I’ll be doing balloon animals near the gift table, which will be under the big oak.” “Balloon animals? Moll, I thought we were professionals.” “A poodle sword is a professional tool if you use it right. At 3:28, you ‘accidentally’ spook the pony they’ve rented. It bolts, you give chase, yelling. The parents will be distracted. That’s my window. I collapse the tablecloth into a sack, presents and all, and toss the bundle over the fence into the azaleas.” Bill stopped sharpening. “You’re banking a lot on a panicky pony.” “It’s a Shetland named Toffee. I’ve been feeding him sugar cubes from the fence line for a week. He’ll bolt if you so much as sneeze. You retrieve the sack, load it into the clown car parked on Maple, and we meet at the rendezvous.” Bill leaned back, the chair groaning in protest. “And the loot? We’re hitting a kid’s party. Are we fencing Lego sets and action figures?” Molly zoomed in on the tablet. “Richard Henderson is paranoid about digital footprints. He buys physical. Rumor has it the big box is the new Hyper-Drive XJ gaming console. Not even on the market yet. The rest is just filler: designer clothes, some solid gold rattle he thinks is ironic, probably a few stock certificates in a shiny envelope. It’s a vanity haul for daddy. We’re just… redistributing his vanity.” “I still don’t like the clown makeup. It itches.” “It’s non-toxic and sweat-resistant. Your tragedy is noted.” Molly locked the tablet. “The real beauty is the cleanup. No one reports the theft until the party’s over. By then, Sparkle and Bungles are a melted memory, and we’re three states away counting gold-plated toy cars.” Bill sheathed his knife. “And if someone, say a particularly observant child, sees you?” Molly stood, stretching. “Then I give them a very special, very twisty balloon giraffe. They’ll be so mesmerized by its existential length, they’ll forget their own name.” She picked up a garish, rainbow-colored wig from a crate. “So. You ready to bring joy to children everywhere, Bungles?” Bill sighed, a long, suffering sound that spoke of diminished dignity. “The things I do for a next-gen console. Let’s go pack the confetti bombs.”
created_at
"2026-02-16T21:36:35.542343+00:00"
2/16/2026, 9:36:35 PM
model
"deepseek/deepseek-v3.2-20251201"
app_id
182717
external_user
(null)
streamed
true
cancelled
false
latency
1533
moderation_latency
(null)
generation_time
52082
tokens_prompt
77
tokens_completion
861
native_tokens_prompt
66
native_tokens_completion
787
native_tokens_completion_images
(null)
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
"stop"
usage
0.00033262
router
(null)
provider_responses
0
id
"728d2214f2894be2b9902b26047e93ae"
status
200
is_byok
false
latency
1533
endpoint_id
"4e9fb7e9-3967-4cd4-9723-df614f9f228e"
provider_name
"Phala"
model_permaslug
"deepseek/deepseek-v3.2-20251201"
api_type
"completions"
id
"gen-1771277742-VXQqI06LDeA34bfnPI5g"
upstream_id
"728d2214f2894be2b9902b26047e93ae"
total_cost
0.00033262
cache_discount
(null)
upstream_inference_cost
0
provider_name
"Phala"
Evaluation details
Result
Evaluator
Details
Meta Data
0.0000%
Matches word count
n/a
neededClean
false
words
540
0.0000%
Dialogue to Total Word Ratio
Ratio: 76.11%, Deviation: 26.11%
neededClean
false
wordsTotal
540
wordsDialogue
411
0.0000%