NC Bench
Overview
About
Tests
Models
Model
Gemini 3 Flash (Preview)
Test
Dialogue tags
Scenario
Write 500 words with 50% dialogue
Run ID
2026-02-08-01KGYWYXYFK70J4PG9BQ7EWN22
Model input
User/Human
You are a author of comedic heist books. Please write a 500 word scene that has 50% 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 neon sign of the Tip-Top Laundromat flickered, casting a sickly green glow over the folding table where Bill had laid out the blueprints. He was sweating despite the air conditioning. It wasn’t the heat; it was the sheer tactical complexity of the inflatable obstacle course. Is that a bouncy house? Molly asked, her eyes narrowing as she pointed a manicured finger at the diagram. It is a fortress of vinyl and air, Bill whispered. It guards the perimeter of the gift table. If we trip the internal pressure sensor, the slide collapses. We’ll be buried in plastic balls before we can reach the LEGO sets. Are you seriously telling me we’re worried about a slide? It’s about the physics, Molly. Do you have the camouflage? Molly reached into her duffel bag and pulled out a bright purple, oversized plush suit with googly eyes the size of dinner plates. I look like a grape that’s had a stroke, she said. Why am I the mascot and you’re the caterer? Because I have the mustache for the mustard stains, Bill replied, smoothing his fake whiskers. Besides, the kids won't suspect a mascot. You represent joy. You represent the spirit of the party. I represent a fire hazard, Bill. This fur is synthetic. One spark from a birthday candle and I’m a violet fireball. What’s the status of the target? The target is a three-tier stack of wrapped boxes situated next to the juice box station, Bill said, tapping a red X on the map. Most of it is fluff. Pajamas, board games, disappointment. But in the center? The Titanium-Edition drone. Six hundred dollars of pure surveillance tech. And the guards? Molly asked. Mothers, Bill said darkly. They operate on a hive mind. They move in clusters of three, armed with iPhones and judgment. If one sees you, they all see you. You have to maintain the character. No matter what they do to you, do not break. What do you mean, what they do to me? Kids are tactile, Molly. They will kick your shins to see if you’re real. They might try to feed you a chicken nugget through the mouth hole. Do not gag. If you gag, we lose the drone. I hate kids, Molly muttered, stroking the mascot’s fuzzy ear. Focus on the prize. When the cake comes out, that’s our window. The sugar rush creates a localized riot. In the confusion, I’ll swap the drone box for a decoy filled with old phone books. You provide the distraction. What kind of distraction? Give them the old Razzle-Dazzle. I am not doing the Razzle-Dazzle in a grape suit. Then we go hungry! Bill slammed his hand on the table. Do you want the drone or do you want your dignity? Molly sighed, looking at the purple head. Fine. But if I get apple juice on my shoes, I’m taking your share of the pawn money.
created_at
"2026-02-08T15:12:18.988716+00:00"
Sunday, February 8th, 2026 at 3:12:18 PM GMT+00:00
model
"google/gemini-3-flash-preview-20251217"
app_id
182717
external_user
(null)
streamed
true
cancelled
false
latency
710
moderation_latency
(null)
generation_time
5684
tokens_prompt
61
tokens_completion
633
native_tokens_prompt
67
native_tokens_completion
654
native_tokens_completion_images
0
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)
origin
"https://ncbench.com/"
is_byok
false
finish_reason
"stop"
native_finish_reason
"STOP"
usage
0.0019955
router
(null)
provider_responses
0
status
200
is_byok
false
latency
710
endpoint_id
"6e82f652-2693-4b60-bbe0-7362d6e345f0"
provider_name
"Google AI Studio"
model_permaslug
"google/gemini-3-flash-preview-20251217"
api_type
"completions"
id
"gen-1770563532-62eU0NJnzgcbgv7PVaNw"
upstream_id
(null)
total_cost
0.0019955
cache_discount
(null)
upstream_inference_cost
0
provider_name
"Google AI Studio"
Evaluation details
Result
Evaluator
Details
Meta Data
95.9868%
Matches word count
n/a
neededClean
false
words
492
0.0000%
Dialogue to Total Word Ratio
Ratio: 0.00%, Deviation: 50.00%
neededClean
false
wordsTotal
492
wordsDialogue
0
47.9934%