NC Bench
Overview
About
Tests
Categories
Models
Vendors
Compare
Run: 2026-04-24-01KQ0672JE2AVBF98ZD4X8NRG7
Run Details
Model:
GPT-OSS 120B
Test:
Dialogue tags
Scenario:
Write 500 words with 30% dialogue
Duration:
3m 43s
Cost:
$0.001179
Provider:
DeepInfra
Evaluation summary
Score
Evaluator
Details Link
0.00%
Matches word count
(details)
35.52%
Dialogue to Total Word Ratio
(details)
17.7601%
Overall
Model input
User/Human
You are a author of comedic heist books. Please write a 500 word scene that has 30% 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 oversized hat that concealed her hair, her eyes scanning the backyard where a rainbow of balloons swayed gently in the early evening breeze. The scent of frosting and cheap plastic glitter clung to the air like a guilty secret. Bill crouched behind the garden gnome, his dark jacket rustling as he shifted his weight. He glanced at his watch, the second hand ticking in rhythm with the distant giggles of children inside the house. A mischievous grin spread across Molly's face. She pulled a crumpled diagram from her pocket, the hastily drawn layout of the party garden, complete with snack tables, a bouncy castle, and the coveted gift pile under the pink canopy. "They're gonna be in there for at least an hour," Bill whispered. "We need a distraction." Molly tapped the side of her chin, considering. "We could rig the sprinklers when they cut the cake. Kids love water fights." "Perfect. While they're scrambling, I'll slip in from the side gate," Bill replied, chuckling. "I'll hop the archway. The clown costume will cover me," Molly said, pulling a red nose from her bag. "Clown outfit?" Bill raised an eyebrow. "Exactly, they'll think I'm part of the show," Molly replied, slipping the nose onto her face. "We'll tell him his superhero powers include moving presents," Molly added, twisting a handful of balloons into a poodle. "Or distract him with a balloon animal. He loves those," Bill suggested. "There." Molly held up the balloon poodle. "Mission accomplished." "Alright, 'Candyland Caper' is a go. Meet at the oak tree in five," Bill instructed. "Got it. No tears, just loot," Molly answered. "The kids will think it's magic," Bill said. "And we walk away with the loot," Molly replied. "I'll secure the presents," she added. "Just keep low," Bill muttered. "Timer set," Molly whispered. "We need a getaway bike," Bill noted. "Got a bike in the shed," Molly responded. "No alarms," Bill declared. "Quiet as mice," Molly agreed. "All set," Bill said. "Let's roll," Molly replied. Molly slipped the balloon poodle into her coat, feeling the soft squeak against her skin. Bill checked his watch, noting the minute hand inching toward midnight. The garden lights flickered, casting long shadows that merged with the whispers of impending mischief. Tonight. for everyone. The night deepened, and the two thieves slipped away from the porch, their footsteps silent on the dew‑covered grass. The party lights flickered, promising a night of chaos and stolen joy.
created_at
"2026-04-24T16:47:22.205Z"
4/24/2026, 4:47:22 PM
model
"openai/gpt-oss-120b"
app_id
182717
external_user
(null)
streamed
true
cancelled
false
latency
321
moderation_latency
(null)
generation_time
222602
tokens_prompt
77
tokens_completion
6221
native_tokens_prompt
114
native_tokens_completion
6184
native_tokens_completion_images
(null)
native_tokens_reasoning
5571
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"
usage
0.001179406
router
(null)
provider_responses
0
endpoint_id
"85835306-dba3-4334-87bc-ab9ffa5c9c62"
id
"chatcmpl-RhEWC0qFVvug0PCcKWy8Zus6"
is_byok
false
latency
99
model_permaslug
"openai/gpt-oss-120b"
provider_name
"DeepInfra"
status
200
user_agent
"langchainjs-openai/1.0.0 ((node/v24.14.1; linux; x64))"
http_referer
(null)
request_id
"req-1777049242-7Up4tVCqDBlBiye8wd2h"
session_id
(null)
api_type
"completions"
id
"gen-1777049242-9JIgNdZspNvsvVB7014O"
upstream_id
"chatcmpl-RhEWC0qFVvug0PCcKWy8Zus6"
total_cost
0.001179406
cache_discount
(null)
upstream_inference_cost
0
provider_name
"DeepInfra"
response_cache_source_id
(null)
Evaluation details
Result
Evaluator
Details
Meta Data
0.00%
Matches word count
n/a
neededClean
false
words
411
35.52%
Dialogue to Total Word Ratio
Ratio: 34.55%, Deviation: 4.55%
neededClean
false
wordsTotal
411
wordsDialogue
142
17.7601%