NC Bench
Overview
About
Tests
Categories
Models
Vendors
Compare
Run: 2026-05-10-01KR912PMEGTYKJMRBPYN2KSBH
Run Details
Model:
Gemma 4 31B
Test:
Dialogue tags
Scenario:
Write 500 words with 30% dialogue
Duration:
19.4s
Cost:
$0.000240
Provider:
DeepInfra
Evaluation summary
Score
Evaluator
Details Link
0.08%
Matches word count
(details)
0.00%
Dialogue to Total Word Ratio
(details)
0.0424%
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
Bill stared at the blueprints of the suburban living room with the intensity of a general planning the invasion of Normandy. He tapped a red marker against a drawing of a beanbag chair. "The perimeter is reinforced with LEGO bricks," Bill whispered, his voice trembling with faux-gravity. "One wrong step and it is a medical emergency." Molly leaned over his shoulder, chewing on a piece of saltwater taffy. She looked at the layout and then at Bill’s tactical gear, which consisted of a black turtleneck and a utility belt holding three different types of glue. "It is a four-year-old's birthday party, Bill," Molly said. "The only security system is a golden retriever with a party hat." "That dog is a sleeper agent," Bill countered, narrowing his eyes. "He knows the scent of greed. We need a diversion. I am thinking a remote-controlled car rigged with a bubble machine to draw the target demographic toward the kitchen." Molly sighed, leaning back in her chair. She began tossing a handful of glitter into the air just to see it float. "Just tell me the extraction point. Do we go through the sliding door or the laundry room?" "The laundry room is too risky," Bill declared, gesturing wildly at the map. "There is a high probability of a stray juice box spill. One slip, and we are sliding across the linoleum like penguins on an oil slick. We go through the patio. We drop from the trellis, neutralize the balloon arch, and swoop in on the gift table." "The gifts are in a pile, Bill," Molly reminded him. "They aren't locked in a vault. They are wrapped in dinosaur paper and held together by scotch tape. We just grab the bag and run." Bill looked offended. "Amateur. We are professionals. We leave no trace except for a single, taunting party blower left at the scene of the crime." Molly stared at him for a long moment. "You spent three hours researching the structural integrity of a bouncy castle, didn't you?" "It was a necessary precaution," Bill snapped. "What if we get trapped in the inflatable abyss? The lack of oxygen alone could be fatal." Molly shook her head and stood up, grabbing her oversized tote bag. "I am going to go get the van. You stay here and make sure your turtleneck doesn't choke you." "Wait," Bill called out as she reached the door. "Did we account for the cake? If there is red velvet, I might be compromised." "Focus, Bill," Molly yelled from the hallway. "The presents first, the frosting later." Bill sighed, adjusted his utility belt, and gave the blueprint one last longing look. He whispered a quiet prayer to the gods of stealth and then accidentally knocked his water glass over the map, erasing the strategic location of the snack table.
created_at
"2026-05-10T13:26:27.737Z"
5/10/2026, 1:26:27 PM
model
"google/gemma-4-31b-it-20260402"
app_id
182717
external_user
(null)
streamed
true
cancelled
false
latency
122
moderation_latency
(null)
generation_time
19259
tokens_prompt
77
tokens_completion
698
native_tokens_prompt
79
native_tokens_completion
605
native_tokens_completion_images
(null)
native_tokens_reasoning
0
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
0.00024017
router
(null)
provider_responses
0
endpoint_id
"4149ab03-3b69-45a0-b94c-6c253b8641a1"
id
"chatcmpl-R50RTq9mY2DBWRnFj2M29Yyq"
is_byok
false
latency
76
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-1778419587-3x1uplU8e4nfnpETGaay"
session_id
(null)
api_type
"completions"
id
"gen-1778419587-OmtePH3AeB5b2dsOHl0w"
upstream_id
"chatcmpl-R50RTq9mY2DBWRnFj2M29Yyq"
total_cost
0.00024017
cache_discount
(null)
upstream_inference_cost
0
provider_name
"DeepInfra"
response_cache_source_id
(null)
Evaluation details
Result
Evaluator
Details
Meta Data
0.08%
Matches word count
n/a
neededClean
false
words
471
0.00%
Dialogue to Total Word Ratio
Ratio: 58.11%, Deviation: 28.11%
neededClean
false
wordsTotal
475
wordsDialogue
276
0.0424%