NC Bench
Overview
About
Tests
Categories
Models
Vendors
Compare
Run: 2026-09-03-01M1KAX2DYV2W53Y4JXN9YA51W
Run Details
Model:
DeepSeek V4 Flash 0731 (Reasoning, High)
Test:
Dialogue tags
Scenario:
Write 500 words with 70% dialogue
Duration:
4m 24s
Cost:
$0.001193
Provider:
OpenInference
Evaluation summary
Score
Evaluator
Details Link
0.00%
Matches word count
(details)
0.07%
Dialogue to Total Word Ratio
(details)
0.0372%
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
The fluorescent bulbs of Molly's kitchen flickered, casting shadows on the blueprint she had taped to the fridge. "Okay, Bill, run the perimeter again," Molly said, tapping the paper. "There's a hedge, a fence, and a woman named Carol who does not take her eyes off the fruit punch," Bill said. He counted on his fingers. "And the target? Timmy. Turning five. Birthday cake at 3 PM, presents at 3:30, meltdown at 3:45," Molly said. "Why is the meltdown part of the plan?" Bill asked. "Because that's our window. When the birthday boy is screaming, all eyes are on the panic," Molly said, grinning. She slid a bag of emergency supplies across the table. Bill peered inside. "Is this a clown wig?" Bill said, holding up the red fluff. "Distraction, Bill. Standard protocol." "I thought standard protocol was a smoke bomb!" "In a bouncy castle? That's a lawsuit. The clown wig is non-toxic and very loud. Clowns make kids run to the corners. Perfect cover." "What if I sneeze? The nose is right there." "Then you sneeze in the direction of the gift table. I'll be doing the 'balloon animal' maneuver. I'll make a poodle. The parents will be watching my craftsmanship." "You can't make a poodle," Bill said. "I've been practicing. Watch. Twist, twist, pinch. That's one leg. Twist, pinch, that's a tail. There. A poodle." Molly held up a deflated balloon shaped vaguely like a constipated worm. "That looks like a sad mangled potato," Bill said. "A potato poodle, Bill. The kids will love it. They'll rush me. While they're rushing, you pull the haul from under the table. We hit the gift table, not the card table." "But what about Timmy's Nana? She's got a walker and she's fast." "Nana's a ringer. She's got coupons for the bouncy castle. She'll use them. Trust me. I've seen her. She tries to bounce on it when the kids are inside." "So we're robbing a family event where a senior citizen commits assault on a bouncy castle?" "We're liberating toys for the black market, Bill. There's a hot trade in pre-screened, parent-approved action figures." Bill slumped in the chair. "What if we get caught?" Bill whined. "What if I trip and the presents fly out and Timmy gets them back and then he cries and I feel bad?" "You will not trip. You will be wearing the 'Mario's Mustache' landing gear. Those rubber soles grip like super glue." "And if Carol from the punch bowl sees my mustache?" "Carol is lactose intolerant. She'll be in the bathroom by 3:32. I've calculated it. She always eats the ice cream cake first." Molly looked at her watch. "We roll in exactly 14 minutes. Let's suit up." Bill sighed. He put on the wig and checked the toaster's reflection. "I look ridiculous." "You look like a master criminal, Bill. A master criminal with a squeaky nose." "And if the master criminal gets attacked by a five-year-old with a lightsaber?" "Then you surrender. 'Timmy, I yield to the power of the Force. Take your presents, young Jedi.'" "That's actually a good plan, Molly." "I know. That's why I'm the brains. Now move, we've got a potato poodle to deliver."
created_at
"2026-09-03T09:51:49.71Z"
9/3/2026, 9:51:49 AM
model
"deepseek/deepseek-v4-flash-20260731"
app_id
182717
external_user
(null)
streamed
true
cancelled
false
latency
566
moderation_latency
(null)
generation_time
262355
tokens_prompt
75
tokens_completion
5836
native_tokens_prompt
67
native_tokens_completion
7437
native_tokens_completion_images
(null)
native_tokens_reasoning
6675
native_tokens_cached
0
num_media_prompt
(null)
num_input_audio_prompt
(null)
num_media_completion
0
num_search_results
(null)
num_fetches
(null)
web_search_engine
(null)
origin
"https://ncbench.com/"
is_byok
false
finish_reason
"stop"
native_finish_reason
"stop"
service_tier
(null)
usage
0.00119327
router
(null)
user_agent
"langchainjs-openai/1.0.0 ((node/v24.19.0; linux; x64))"
http_referer
(null)
request_id
"req-1788429109-3DTk40H689Z1Otk6G23k"
session_id
(null)
preset_id
(null)
api_type
"completions"
id
"gen-1788429109-IRHXnu3o3dXgTAkP5XJg"
upstream_id
"chatcmpl-ad132e8e-aa3d-47ee-b21c-f430d12ab76c"
provider_responses
0
endpoint_id
"14c4df90-c7af-42ff-99e7-69d62a89e7d5"
is_byok
false
latency
1123
model_permaslug
"deepseek/deepseek-v4-flash-20260731"
provider_name
"AkashML"
status
429
1
endpoint_id
"f954a48e-1c90-433b-9348-4720f8030331"
id
"chatcmpl-ad132e8e-aa3d-47ee-b21c-f430d12ab76c"
is_byok
false
latency
566
model_permaslug
"deepseek/deepseek-v4-flash-20260731"
provider_name
"OpenInference"
status
200
total_cost
0.00119327
cache_discount
(null)
upstream_inference_cost
0
provider_name
"OpenInference"
response_cache_source_id
(null)
data_region
"global"
workspace_id
"97e315e5-d303-487d-83c1-83180e8a13d4"
Evaluation details
Result
Evaluator
Details
Meta Data
0.00%
Matches word count
n/a
neededClean
false
words
535
0.07%
Dialogue to Total Word Ratio
Ratio: 82.00%, Deviation: 12.00%
neededClean
false
wordsTotal
539
wordsDialogue
442
0.0372%