NC Bench
Overview
About
Tests
Categories
Models
Vendors
Compare
Run: 2026-07-25-01KYCT86M4W6HN9KGWXPH2EVFW
Run Details
Model:
Gemini 3.6 Flash (Reasoning)
Test:
Dialogue tags
Scenario:
Write 500 words with 30% dialogue
Duration:
35.0s
Cost:
$0.063153
Provider:
Google AI Studio
Evaluation summary
Score
Evaluator
Details Link
68.10%
Matches word count
(details)
52.57%
Dialogue to Total Word Ratio
(details)
60.3344%
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 unrolled the blueprint of forty-two Elm Street across the greasy basement workbench, pinning the curled corners down with two half-eaten corn dogs. Bill leaned in close, squinting through his tactical night-vision goggles despite the buzzing fluorescent light humming overhead. The target was formidable: Little Timmy’s seventh birthday bash. Intel suggested an untapped goldmine of unopened Lego sets, a shiny drone, and six mystery boxes wrapped in metallic foil. Security was exceptionally tight—four hyper-vigilant mothers, an overworked party magician, and twenty sugar-crazed second graders armed with noisy blowers. The basement smelled faintly of damp cardboard and stolen industrial-grade bubble solution. A single floodlight illuminated their makeshift tactical table, highlighting complex entry trajectories sketched in neon crayon. Molly picked up a magnifying glass, scrutinizing the patio perimeter. The perimeter fence was six feet of solid cedar, easily scaled, but the true threat lurked near the sandbox. "We hit the bouncy castle at fourteen hundred hours," Molly whispered, tapping her purple marker on the backyard schematic. "Is the bouncy castle our insertion point or the getaway vehicle?" Bill asked, adjusting the collar of his turtleneck. "Because last time we used an inflatable, my leg got stuck in the turret." Molly rubbed her temples, leaving a purple streak across her forehead. The tactical implications of an inflatable fortress were complex. If inflated, it provided optimal camouflage. If deflated, it became a suffocating vinyl trap. "It’s the entry vector, Bill," Molly said, rolling her eyes. "We slice through the rear nylon, bypass the naptime guards, and bag the primary loot." "What about the clown?" Bill inquired nervously. "Is Bongo still on duty?" "Bongo is an unpredictable tactical wildcard," Molly admitted softly. "He balloon-twists on sight. If he approaches, do not engage. Avoid his blind spot at all costs." Bill checked his utility belt. It was stuffed with smoke pellets, duct tape, and four organic juice boxes to bribe stray toddler sentries. He proudly pulled out a pair of bright pink, glittery novelty glasses. "Are these for tactical camouflage?" Bill asked. "Put those away before I throw you in the sandbox," Molly snapped. "Focus. Are the sugar levels optimal?" "Cake cutting is scheduled for precisely two-fifteen," Bill reported, checking his stolen smartwatch. "The targets will enter a profound, frosting-induced coma shortly after." "That gives us a clean seven-minute operational window," Molly said, nodding in grim satisfaction. "Are we really going after the main Lego Star Wars set?" Bill whispered in genuine awe. "We are taking everything, Bill," Molly declared, leaning over the table. "The Legos, the remote-controlled cars, and if time permits, every goodie bag on that folding table." "Even the small plastic whistles?" Bill gasped. "Especially the whistles," Molly replied coldly. Molly pulled down her knitted ski mask, which featured an embroidered kitten because she had bought it at a craft fair. Bill nodded solemnly, locking a fresh canister of industrial bubble solution into his plastic blaster. They synchronized their cheap digital watches, hoisted their heavy canvas sacks, and stepped into the dark alleyway, completely ready to strip a spoiled seven-year-old of his absolute entire birthday fortune.
created_at
"2026-07-25T14:18:16.59Z"
7/25/2026, 2:18:16 PM
model
"google/gemini-3.6-flash-20260721"
app_id
182717
external_user
(null)
streamed
true
cancelled
false
latency
1421
moderation_latency
(null)
generation_time
34936
tokens_prompt
75
tokens_completion
3385
native_tokens_prompt
67
native_tokens_completion
8407
native_tokens_completion_images
0
native_tokens_reasoning
7704
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
"default"
usage
0.063153
router
(null)
provider_responses
0
endpoint_id
"5d6d133d-a953-4781-a835-7ee76f4e1388"
id
"qMVkaoq7KLOA-8YP3v3h4QU"
is_byok
false
latency
1421
model_permaslug
"google/gemini-3.6-flash-20260721"
provider_name
"Google AI Studio"
status
200
user_agent
"langchainjs-openai/1.0.0 ((node/v24.18.0; linux; x64))"
http_referer
(null)
request_id
"req-1784989096-59CcmDCG6S9wI1azKsS3"
session_id
(null)
preset_id
(null)
api_type
"completions"
id
"gen-1784989096-kuYvKHbCK3dMzOipS397"
upstream_id
"qMVkaoq7KLOA-8YP3v3h4QU"
total_cost
0.063153
cache_discount
(null)
upstream_inference_cost
0
provider_name
"Google AI Studio"
response_cache_source_id
(null)
data_region
"global"
Evaluation details
Result
Evaluator
Details
Meta Data
68.10%
Matches word count
n/a
neededClean
false
words
514
52.57%
Dialogue to Total Word Ratio
Ratio: 33.59%, Deviation: 3.59%
neededClean
false
wordsTotal
527
wordsDialogue
177
60.3344%