NC Bench
Overview
About
Tests
Categories
Models
Vendors
Compare
Run: 2026-10-20-01M5C68ZEXTW4BV40WQQXYPQBT
Run Details
Model:
DeepSeek V4.1 Flash 1017 (Reasoning, High)
Test:
Dialogue tags
Scenario:
Write 500 words with 50% dialogue
Duration:
51.5s
Cost:
$0.004365
Provider:
InferenceNet
Evaluation summary
Score
Evaluator
Details Link
0.00%
Matches word count
(details)
0.52%
Dialogue to Total Word Ratio
(details)
0.2576%
Overall
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
Molly adjusted the van’s mirror and watched tiny pirates and princesses funnel through a door decorated with balloon dinosaurs. A bouncy castle wheezed in the backyard. “Remind me why we’re robbing a six-year-old’s birthday party instead of a bank?” Bill lowered his binoculars. “Banks have guards, cameras, and consequences. Birthday parties have cake.” “They also have parents.” “Slow parents. Tired parents. Parents who’ve handed out juice boxes and lost the will to police anything.” Molly counted the gifts stacked beside the bouncy castle: three boxes, one gift bag, and something shaped suspiciously like a television. “And the entry?” Bill produced a limp clown suit from a duffel bag. “I go in as Mr. Wiggles. You’re my assistant, Mrs. Wiggles.” “I am not being Mrs. Wiggles.” “Then you can be the birthday magician.” “I don’t do magic.” “You made our landlord believe the rent was coming. That’s magic.” Molly rubbed her forehead and watched a father in a superhero cape stagger past with a giant gift box. “What about the birthday kid?” “We leave one present. A nice one. A book.” “A book?” “Books are educational. We’re not monsters.” “We’re stealing from a six-year-old.” “We’re redistributing joy. Efficiently.” Bill tapped the dashboard as a mother carried out a cake. “Simple plan. You trip over the clown. I grab the pile. We leave before the singing starts.” “That’s your plan?” “It worked at the wedding.” “We got banned from three counties.” “We got banned from three counties and kept the blender.” Molly pulled on gloves. “What if they sing?” “Then we sing too. Blend in. Happy birthday, tiny mark.” “You’re going to hell, Bill.” “Probably. But I’ll be early, and I’ll bring a gift bag.” The party had spread into the garden: a piñata shaped like a llama, a table of cupcakes, and a hired entertainer warming up a puppet. Molly noted two exits, one hedge, and a dog that looked suspiciously alert. “The dog’s a problem,” she said. Bill squinted. “That’s a golden retriever.” “That’s a witness.” “He’s wearing a party hat.” “Exactly. He’s undercover.” Molly watched the entertainer produce a rabbit from a hat. “We could use him. A real magician creates chaos.” “Or he creates witnesses,” Bill said. “Stick to the clown. Clowns are invisible at children’s parties.” “No one trusts a clown.” “Exactly. They’re too busy judging.” A child in a tutu pointed at the van. Molly ducked. “He’s pointing.” “He’s waving,” Bill said. “Waving is pointing with optimism.” “Then we’re doomed.” Molly stared at the clown suit, then at the pile of presents. “One present. We leave one present.” “The book.” “The book.” Bill grinned and started the van. “Mrs. Wiggles, let’s go ruin a childhood.”
created_at
"2026-10-20T08:19:03.267Z"
10/20/2026, 8:19:03 AM
model
"deepseek/deepseek-v4.1-flash-20261017"
app_id
182717
external_user
(null)
streamed
true
cancelled
false
latency
125
moderation_latency
(null)
generation_time
51495
tokens_prompt
75
tokens_completion
8028
native_tokens_prompt
67
native_tokens_completion
8719
native_tokens_completion_images
(null)
native_tokens_reasoning
8021
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
0.00436486
router
(null)
user_agent
"langchainjs-openai/1.0.0 ((node/v24.21.0; linux; x64))"
http_referer
(null)
request_id
"req-1792484343-pPHp0jK5Kb8h3SpPLNeE"
session_id
(null)
preset_id
(null)
api_type
"completions"
id
"gen-1792484343-EEiWFItxZ5yR87735N6v"
upstream_id
"gv2_xyjHjgCzUxjuzk4IqTczV"
provider_responses
0
endpoint_id
"4231aea9-73af-4a2f-bd14-660b1d8b13b2"
id
"gv2_xyjHjgCzUxjuzk4IqTczV"
is_byok
false
latency
125
model_permaslug
"deepseek/deepseek-v4.1-flash-20261017"
provider_name
"InferenceNet"
status
200
total_cost
0.00436486
cache_discount
(null)
upstream_inference_cost
0
provider_name
"InferenceNet"
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
449
0.52%
Dialogue to Total Word Ratio
Ratio: 60.26%, Deviation: 10.26%
neededClean
false
wordsTotal
453
wordsDialogue
273
0.2576%