NC Bench
Overview
About
Tests
Categories
Models
Vendors
Compare
Run: 2026-05-10-01KR8VEW6X8DRNWP51B5H75RSS
Run Details
Model:
Gemma 4 26B
Test:
Dialogue tags
Scenario:
Write 500 words with 70% dialogue
Duration:
1m 54s
Cost:
$0.000207
Provider:
DekaLLM
Evaluation summary
Score
Evaluator
Details Link
27.17%
Matches word count
(details)
0.00%
Dialogue to Total Word Ratio
(details)
13.5829%
Overall
Model input
User/Human
You are a author of comedic heist books. Please write a 500 word scene that has 70% 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 leaned over the blueprints, which were actually just a crudely drawn map of a suburban backyard on a cocktail napkin. Bill stared at the napkin with profound suspicion. Is this a napkin, Molly? It is a strategic tactical layout, Bill. It is a napkin from a Mexican restaurant. There is a salsa stain where the bouncy castle is supposed to be. That is a high-intensity zone, Bill. Don’t be pedantic. The target is a six-year-old’s birthday party. We are professional thieves. We once bypassed a laser grid in a Zurich museum. We also almost died because you forgot the glass cutter, Molly. Focus. The objective is the gift pile. It is located in the sunroom, guarded by a mountain of colorful wrapping paper and a very distracted toddler in a dinosaur costume. A dinosaur? Is he armed? He is unpredictable. He has sticky hands and zero respect for personal space. So, what is the entry point? The sliding glass door. But it is rigged. Rigged how? With pressure sensors? With a motion-activated bubble machine. If we trigger it, we will be covered in soapy spheres, and the visual interference will be catastrophic. So we bring umbrellas? No, Bill. We bring a sponge. We soak ourselves, we blend in with the chaos, and we move during the cake cutting. The cake cutting is the perfect window of vulnerability. Exactly. While the adults are distracted by sugar crashes and singing, we execute the sweep. How many presents are we talking about? Based on the guest list, approximately twenty-four. High-value targets include LEGO sets, remote-controlled drones, and possibly a high-end tablet. And the getaway vehicle? The minivan. I have it parked two blocks away, disguised as a generic florist van. A florist van? Why a florist? Because no one suspects a man in a floral apron of being a master infiltrator. You look ridiculous in that apron, Molly. It is a disguise, Bill! It provides psychological cover! Do we have the bags? The reinforced duffel bags are in the trunk. They are designed to hold heavy objects without making a suspicious clinking sound. Good. And the distraction? I will initiate the distraction by pretending to be a lost uncle who is very interested in the finger foods. A lost uncle? That is your master plan? I will blend in. I will ask about the trajectory of the bouncy castle. I will be invisible. You are about as invisible as a wrecking ball, Bill. Just stay low, keep your eyes on the dinosaur, and for heaven's sake, do not eat any of the cupcakes. Why not? Because once you start on the buttercream, the mission is over. You lose all tactical discipline. Fine. Let's go steal some toys. Wait, did you bring the extra batteries for the drone? I thought you brought the batteries! This is why we are not allowed in Zurich anymore!
created_at
"2026-05-10T11:48:15.34Z"
5/10/2026, 11:48:15 AM
model
"google/gemma-4-26b-a4b-it-20260403"
app_id
182717
external_user
(null)
streamed
true
cancelled
false
latency
1707
moderation_latency
(null)
generation_time
113728
tokens_prompt
77
tokens_completion
761
native_tokens_prompt
68
native_tokens_completion
614
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
0.0002067
router
(null)
provider_responses
0
endpoint_id
"2e9daaa4-87bf-49f7-a44a-2c1d617c2c51"
is_byok
false
latency
92
model_permaslug
"google/gemma-4-26b-a4b-it-20260403"
provider_name
"Venice"
status
429
1
endpoint_id
"d99bcbe6-bec3-4991-965e-45fea79fe63c"
id
"chatcmpl-9954aea3e4f0b94c"
is_byok
false
latency
1707
model_permaslug
"google/gemma-4-26b-a4b-it-20260403"
provider_name
"DekaLLM"
status
200
user_agent
"langchainjs-openai/1.0.0 ((node/v24.14.1; linux; x64))"
http_referer
(null)
request_id
"req-1778413695-otE9nTZTFbDvwGHAksee"
session_id
(null)
api_type
"completions"
id
"gen-1778413695-Krgnt6l6hhTwzUWyX0RY"
upstream_id
"chatcmpl-9954aea3e4f0b94c"
total_cost
0.0002067
cache_discount
(null)
upstream_inference_cost
0
provider_name
"DekaLLM"
response_cache_source_id
(null)
Evaluation details
Result
Evaluator
Details
Meta Data
27.17%
Matches word count
n/a
neededClean
false
words
481
0.00%
Dialogue to Total Word Ratio
Ratio: 0.00%, Deviation: 70.00%
neededClean
false
wordsTotal
489
wordsDialogue
0
13.5829%