[mle]/DaBench

GPT-6 Luna vs Qwen3.8 27B

Side A
Model page
Side B
Model page
Overall
61
vs
Overall
48

GPT-6 Luna is clearly better. It wins 51 of the 68 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. Qwen3.8 27B spends 3.9× as many thinking tokens per test case.

Thinking tokensAnswer tokens
TaskAB
Overall
6149.1s
1.9K, 73% thinking5.9K, 91% thinking
4833.3s
Agents
561m 18s
1.3K, 67% thinking5.2K, 78% thinking
5736.6s
Structured Extraction
9514.5s
999, 72% thinking4.6K, 92% thinking
7320.6s
Coding
732m 18s
3.6K, 60% thinking9.3K, 94% thinking
3753.0s
Grounded Answers
6715.6s
1.1K, 72% thinking4.2K, 91% thinking
6131.6s
Instruction Following
8432.3s
2K, 74% thinking6.9K, 93% thinking
6733.6s
Business Logic
3743.1s
3.2K, 93% thinking7.6K, 98% thinking
3632.9s
Security & permissions
6224.5s
1K, 68% thinking3.7K, 85% thinking
5122.6s
Diagnosis & investigation
1737.7s
1.1K, 68% thinking5.9K, 85% thinking
040.0s

What you'd run

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

SpecAB
Speed
90 tok/s
196 tok/s (better)
Machine
Hosted
Hosted
Memory at 32K
—
—
Context
1.1M (better)
256K
Cost per run
$0.10 (better)
$1.28
Cost per case
$0.00067 (better)
$0.0083
Time per case
49.1s
33.3s (better)
Quantization
—
FP8
Harness
openrouter
openrouter

The race

2 lead changes

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

GPT-6 LunaQwen3.8 27B
025507510050100150Perfect run6148

Every test case

153 test cases

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

51 won by GPT-6 Luna17 won by Qwen3.8 27B85 tied