NC Bench
Overview
About
Tests
Categories
Models
Vendors
Compare
Run: 2026-05-10-01KR92RRFAD4A2NMS5T5PMW3CH
Run Details
Model:
Gemma 4 26B (Reasoning)
Test:
Dialogue tags
Scenario:
Write 500 words with 30% dialogue
Duration:
9m 5s
Cost:
$0.004869
Provider:
Novita
Evaluation summary
Score
Evaluator
Details Link
0.00%
Matches word count
(details)
99.18%
Dialogue to Total Word Ratio
(details)
49.5879%
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
The interior of the rusted transit van smelled of stale coffee, empty snack wrappers, and the heavy, lingering scent of nervous sweat. Outside, the quiet suburban street was bathed in the pale, mocking glow of streetlights that illuminated perfectly manicured lawns and silent, sleeping houses. Molly spread a crinkled, glitter-flecked piece of cardstock across the cluttered dashboard, illuminating the target with a focused penlight. It was not a high-security bank vault blueprint, but to her, the stakes were just as high. She pointed at a crayon-drawn sun with the grim precision of a general marking a battlefield, her eyes scanning the colorful, scented perimeter of the birthday invitation. The shadows in the van danced against the ceiling as the engine idled with a rhythmic, low-frequency thrum, and the smell of diesel and anticipation hung heavy in the cramped space. "Listen, Bill. One mistake and we face a sugar-fueled riot," Molly whispered, her voice dropping into a low, professional register. Bill leaned in, squinting at the invitation while adjusting his tactical vest, which was struggling to contain his midsection. He shifted uncomfortably on the cracked vinyl seat, the springs squeaking in protest of his movement. "Are those dinosaur stickers? I need them for morale. I also have my specialized grape juice extractor ready." Molly hissed a warning and tapped the center of the paper. "Focus. The target is the gift pile in the foyer, situated between the juice boxes and the bounce house. We move during the singing of Happy Birthday to avoid the Toddler Surge. Once the cake is cut, the kinetic energy in that room will be off the charts." Bill wiped sweat from his brow, his eyes darting toward the brightly lit house where muffled, joyous laughter drifted through the air. He seemed less concerned with the loot and more with the various atmospheric hazards of the impending party, his hands twitching rhythmically against his thighs. "What about the clown? I've heard they have telescopic reach. And balloon animals could be weapons." "The clown is a variable," Molly admitted, checking her heavy-duty waterproof watch. "We will bypass the sandbox perimeter and secure the LEGO sets before the first meltdown occurs. No distractions, no clowns, and absolutely no magicians. Just move." Bill reached for his lockpicks, his expression hardening into a look of faux-intensity that barely masked his trembling fingers. The metal tools clinked softly in their pouch, a tiny, sharp sound in the heavy silence of the van. "Understood. Stealth, speed, and no eye contact with the juice-box brigade. Copy that." Molly gripped the steering wheel, her face set in grim determination as the first notes of a distant, out-of-tune nursery rhyme drifted through the van's thin walls. The heist was officially on. "Let's get those presents."
created_at
"2026-05-10T13:55:59.105Z"
5/10/2026, 1:55:59 PM
model
"google/gemma-4-26b-a4b-it-20260403"
app_id
182717
external_user
(null)
streamed
true
cancelled
false
latency
363
moderation_latency
(null)
generation_time
534400
tokens_prompt
77
tokens_completion
12858
native_tokens_prompt
82
native_tokens_completion
12147
native_tokens_completion_images
(null)
native_tokens_reasoning
12131
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
0.00486946
router
(null)
provider_responses
0
endpoint_id
"eb382d64-dce1-42d5-b23d-6fefb91948a3"
is_byok
false
latency
10077
model_permaslug
"google/gemma-4-26b-a4b-it-20260403"
provider_name
"SiliconFlow"
status
504
1
endpoint_id
"03b96766-ef7e-454f-9519-c76b70654b49"
id
"9283c76323c27eae07d1a9638b5d2fd0"
is_byok
false
latency
287
model_permaslug
"google/gemma-4-26b-a4b-it-20260403"
provider_name
"Novita"
status
200
user_agent
"langchainjs-openai/1.0.0 ((node/v24.14.1; linux; x64))"
http_referer
(null)
request_id
"req-1778421359-OJ6eoni0atw1IALdu9le"
session_id
(null)
api_type
"completions"
id
"gen-1778421359-l6Q5BrJwSVCGt7wObtb7"
upstream_id
"9283c76323c27eae07d1a9638b5d2fd0"
total_cost
0.00486946
cache_discount
(null)
upstream_inference_cost
0
provider_name
"Novita"
response_cache_source_id
(null)
Evaluation details
Result
Evaluator
Details
Meta Data
0.00%
Matches word count
n/a
neededClean
false
words
459
99.18%
Dialogue to Total Word Ratio
Ratio: 30.41%, Deviation: 0.41%
neededClean
false
wordsTotal
467
wordsDialogue
142
49.5879%