[mle]/DaBench

GPT-6 Luna vs Spark X2.6 4B

Side A
Model page
Side B
Model page
Overall
61
vs
Overall
15

GPT-6 Luna is clearly better. It wins 87 of the 90 test cases where they differ.

Learn to build evals like this in the AI Engineering Academy

Task by task

A score turns orange only when that model wins clearly more of the task's test cases. Under each score, the average time per test case on the machine it ran on. Spark X2.6 4B spends 5.0× as many thinking tokens per test case.

Thinking tokensAnswer tokens
TaskAB
Overall
6149.1s
1.9K, 73% thinking8.2K, 84% thinking
151m 3s
Agents
561m 18s
1.3K, 67% thinking11.5K, 97% thinking
141m 32s
Structured Extraction
9514.5s
999, 72% thinking5.3K, 78% thinking
1039.5s
Coding
732m 18s
3.6K, 60% thinking9.1K, 48% thinking
41m 8s
Grounded Answers
6715.6s
1.1K, 72% thinking6.6K, 86% thinking
1050.2s
Instruction Following
8432.3s
2K, 74% thinking11.9K, 90% thinking
371m 29s
Business Logic
3743.1s
3.2K, 93% thinking4.9K, 86% thinking
1036.3s
Security & permissions
6224.5s
1K, 68% thinking7K, 94% thinking
3154.3s
Diagnosis & investigation
1737.7s
1.1K, 68% thinking11.6K, 98% thinking
01m 30s

What you'd run

The build behind each score. Orange marks the faster, smaller, cheaper or longer-context side.

SpecAB
Speed
90 tok/s
136 tok/s
Machine
Hosted
RTX 5090, 32 GB
Memory at 32K
—
9.4 GB
Context
1.1M (better)
605K
Cost per run
$0.10 (better)
$1.61 est.
Cost per case
$0.00067 (better)
$0.011 est.
Time per case
49.1s (better)
1m 3s
Quantization
—
BF16
Harness
openrouter
llama.cpp

The race

No lead changes

Each model's running score as the test cases go by. The dashed line is a perfect run.

GPT-6 LunaSpark X2.6 4B
0255075100501001506115

Every test case

153 test cases

Each square is one test case, colored by the model that scored higher.

87 won by GPT-6 Luna3 won by Spark X2.6 4B63 tied