NC Bench
Overview
About
Tests
Categories
Models
Vendors
Compare
Run: 2026-04-27-01KQ7265TEKCDT3B5DRMQ2Y874
Run Details
Model:
Qwen 3.6 35B
Test:
Dialogue tags
Scenario:
Write 500 words with 30% dialogue
Duration:
48.4s
Cost:
$0.009166
Provider:
AtlasCloud
Evaluation summary
Score
Evaluator
Details Link
0.00%
Matches word count
(details)
99.99%
Dialogue to Total Word Ratio
(details)
49.9926%
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 fluorescent bulb in the back of the stolen delivery van buzzed like an angry hornet. Molly unrolled a laminated floor plan of the Henderson residence across her knees, smoothing out a sticky residue with the heel of her glove. Bill sat opposite her, meticulously wrapping a roll of electrical tape around his wrists. They were professionals, or at least they had been until the pension fund collapsed and someone suggested a simpler target. The rain hammered against the metal roof, drowning out the distant sound of traffic and normal human ambition. Inside the cramped cabin, the smell of wet canvas and cheap coffee hung heavy in the stagnant air. Bill adjusted his cracked leather gloves and stared at the blueprints like they were ancient maps to a forgotten kingdom. He had spent the last decade perfecting lock picking algorithms and bypassing laser grids, yet here he was, planning to raid a suburban birthday party. The logistics were absurd, but the payout was substantial, and neither of them had dignity left to spare. The van smelled of damp wool and old fast food. Molly smoothed the creases on the map with deliberate, practiced motions. Bill watched her hands, noting the scar running along his thumb. They had robbed armored trucks and jewelry vaults, but this operation felt different. The stakes were lower, the target was softer, and the success rate should have been impossibly high. They adjusted their breathing and synchronized their watches. The street outside was quiet, save for the occasional passing car and the distant barking of a neighborhood dog. Molly packed the zip ties into her coat pocket and checked the weight of the dolly straps. Bill tested the tension on the window cutter blade. They were ready. The plan was flawless. The execution would be simple. They just needed to walk into the party, ignore the screaming children, and walk out with every box. "Thirty-seven gifts," Bill muttered. "Mostly electronics." Molly tapped the paper. "The patio door stays open. I tested the latch last Tuesday. It swings with the reluctance of a toddler denied candy." She flipped the page. "The camera is fake. Just a plastic shell wired to a wind chime. We enter between piƱata and cake." Bill nodded slowly, reaching for a digital timer. "If the boy opens a dinosaur first, the noise triggers panic. Panic brings parents. Parents bring authorities. Authorities bring jail time." "We use balloons," Molly replied. "Glitter filled. Popped simultaneously. Creates overwhelming cuteness. No adult resists confetti. They will lower the toy. We bag the loot and exit through the garage. The dolly has shock absorbers." Bill adjusted his fogged glasses. "Receipt folders? Mothers always keep them." "We relocate assets. Do not return anything." She tossed him a burlap sack. "Forty minutes until punch is served. Pack the coffee. We move when the candles blow out." Bill caught the bag. "I handle cutters. You take ties. We are ghosts. Gentle ghosts." "Check timers," Molly said, staring at the green numbers. "I hope the cake is moist. Dry cake means stressed parents. Stressed parents mean missed receipts. Missed receipts mean full payout." Molly smiled, tracing the living room wall. "Let them forget." The engine hummed softly as they prepared for the final approach. Rain continued to fall on the suburban streets.
created_at
"2026-04-27T08:51:39.477Z"
4/27/2026, 8:51:39 AM
model
"qwen/qwen3.6-35b-a3b-20260415"
app_id
182717
external_user
(null)
streamed
true
cancelled
false
latency
997
moderation_latency
(null)
generation_time
48199
tokens_prompt
77
tokens_completion
10007
native_tokens_prompt
76
native_tokens_completion
9483
native_tokens_completion_images
(null)
native_tokens_reasoning
8790
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
0.009165716
router
(null)
provider_responses
0
endpoint_id
"179c6985-3a1c-48f1-a009-9f3e3e402cdd"
id
"e914935393dd4d5398514cf67b61e091"
is_byok
false
latency
997
model_permaslug
"qwen/qwen3.6-35b-a3b-20260415"
provider_name
"AtlasCloud"
status
200
user_agent
"langchainjs-openai/1.0.0 ((node/v24.14.1; linux; x64))"
http_referer
(null)
request_id
"req-1777279899-YVnQjjZNSD80poxdiDAQ"
session_id
(null)
api_type
"completions"
id
"gen-1777279899-FnLVAxsFfZQEHEyZ5FR6"
upstream_id
"e914935393dd4d5398514cf67b61e091"
total_cost
0.009165716
cache_discount
(null)
upstream_inference_cost
0
provider_name
"AtlasCloud"
response_cache_source_id
(null)
Evaluation details
Result
Evaluator
Details
Meta Data
0.00%
Matches word count
n/a
neededClean
false
words
550
99.99%
Dialogue to Total Word Ratio
Ratio: 29.95%, Deviation: 0.05%
neededClean
false
wordsTotal
551
wordsDialogue
165
49.9926%