NC Bench
Overview
About
Tests
Categories
Models
Vendors
Compare
Run: 2026-05-10-01KR90M89ED5EVHWZPGQAY4YVA
Run Details
Model:
Gemma 4 31B (Reasoning)
Test:
Dialogue tags
Scenario:
Write 500 words with 30% dialogue
Duration:
3m 32s
Cost:
$0.002089
Provider:
DeepInfra
Evaluation summary
Score
Evaluator
Details Link
75.16%
Matches word count
(details)
76.46%
Dialogue to Total Word Ratio
(details)
75.8063%
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 spread the blueprint across the card table, weighting the corners with stale donuts. The schematic of the Miller residence was detailed, marking the exact trajectory from the patio to the gift table. Bill stared at the drawing, squinting through a pair of oversized binoculars he did not need because they were indoors. "Why the perimeter check?" Bill asked. "It is a bunch of second graders, Molly." Molly slammed her hand on the table, sending a sprinkle flying. She viewed the chaos of a children's party not as a celebration, but as a tactical nightmare. The noise levels alone would mask their entry, but the unpredictable nature of sugar-high toddlers was a variable she could not quantify. She pointed to a red X near the cake. "Sugar-high kids are the most dangerous operatives on the planet," she hissed. "One tantrum and the whole operation is compromised." Bill shrugged, leaning back in his creaky folding chair. He was more interested in the potential haul. The guests were from the upper crust of the suburbs, meaning the presents were not just plastic whistles; they were high-ticket electronics and designer plushies. He imagined the resale value of a limited-edition robotic dinosaur. "I can handle the kids," Bill said. "I will just distract them with a magic trick or something." "The cake is the trigger," Molly countered. "When the candles go out, we move." "What if they have a bouncy house?" Bill asked. "I am not going in a bouncy house." "The bouncy house is a diversion, you idiot," she snapped. "We stay low, we stay fast, and we leave no trace." "I am just saying, if there is a bouncy house, I might want to jump in it first." "Will there be pizza?" Bill asked. "I am not doing a heist on an empty stomach." Molly sighed, rubbing her temples. Bill's idea of a magic trick usually involved accidentally gluing his fingers together. She shifted her focus to the timing. The window of opportunity opened exactly ten minutes after the cake was served, while the adults were distracted by the inevitable frosting-based brawl. "You enter via the laundry chute," Molly commanded. "I will provide cover by pretending to be a distant cousin from Ohio." Bill frowned, considering the logistics of the chute. He was a large man, built like a refrigerator with a grudge. Sliding down a plastic tube sounded like a very tight squeeze. He checked his watch and then looked at the blueprint again, tracing the line to the gift table. "Do I get to wear the costume?" he asked. Molly pulled a shimmering, oversized clown suit from a garment bag. It was neon yellow with polka dots that hurt the eyes. She looked at Bill, then at the suit, and finally at the blueprints. The plan was absurd, the target was ridiculous, but the payoff was promising. "Only if you can keep your mouth shut," she replied.
created_at
"2026-05-10T13:18:34.295Z"
5/10/2026, 1:18:34 PM
model
"google/gemma-4-31b-it-20260402"
app_id
182717
external_user
(null)
streamed
true
cancelled
false
latency
172
moderation_latency
(null)
generation_time
211849
tokens_prompt
77
tokens_completion
5580
native_tokens_prompt
82
native_tokens_completion
5469
native_tokens_completion_images
(null)
native_tokens_reasoning
4847
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
0.00208888
router
(null)
provider_responses
0
endpoint_id
"4149ab03-3b69-45a0-b94c-6c253b8641a1"
id
"chatcmpl-RMQLch7ll7MXKulJvVAHsF7U"
is_byok
false
latency
47
model_permaslug
"google/gemma-4-31b-it-20260402"
provider_name
"DeepInfra"
status
200
user_agent
"langchainjs-openai/1.0.0 ((node/v24.14.1; linux; x64))"
http_referer
(null)
request_id
"req-1778419114-x2XgrskfNrkQ3gXeH9XY"
session_id
(null)
api_type
"completions"
id
"gen-1778419114-knxZKzVgfz3nhnXeryba"
upstream_id
"chatcmpl-RMQLch7ll7MXKulJvVAHsF7U"
total_cost
0.00208888
cache_discount
(null)
upstream_inference_cost
0
provider_name
"DeepInfra"
response_cache_source_id
(null)
Evaluation details
Result
Evaluator
Details
Meta Data
75.16%
Matches word count
n/a
neededClean
false
words
487
76.46%
Dialogue to Total Word Ratio
Ratio: 32.32%, Deviation: 2.32%
neededClean
false
wordsTotal
492
wordsDialogue
159
75.8063%