[mle]/DaBench

GPT-6 Luna vs Muse Spark 1.3 Contributor

Side A
Model page
Side B
Model page
Overall
61
vs
Overall
66

Too close to call. Muse Spark 1.3 Contributor wins 25 of the 40 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. Muse Spark 1.3 Contributor spends 5.8× as many thinking tokens per test case.

Thinking tokensAnswer tokens
TaskAB
Overall
6149.1s
1.9K, 73% thinking8.7K, 91% thinking
661m 42s
Agents
561m 18s
1.3K, 67% thinking14.7K, 88% thinking
573m 16s
Structured Extraction
9514.5s
999, 72% thinking4.2K, 91% thinking
9037.9s
Coding
732m 18s
3.6K, 60% thinking10.8K, 95% thinking
482m 1s
Grounded Answers
6715.6s
1.1K, 72% thinking6.1K, 91% thinking
811m 1s
Instruction Following
8432.3s
2K, 74% thinking8.7K, 91% thinking
841m 18s
Business Logic
3743.1s
3.2K, 93% thinking6K, 95% thinking
761m 7s
Security & permissions
6224.5s
1K, 68% thinking8.2K, 87% thinking
761m 59s
Diagnosis & investigation
1737.7s
1.1K, 68% thinking17.2K, 90% thinking
173m 34s

What you'd run

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

SpecAB
Speed
90 tok/s
107 tok/s (better)
Machine
Hosted
Hosted
Memory at 32K
—
—
Context
1.1M (better)
1M
Cost per run
$0.10 (better)
$0.41
Cost per case
$0.00067 (better)
$0.0027
Time per case
49.1s (better)
1m 42s
Quantization
—
—
Harness
openrouter
openrouter

The race

5 lead changes

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

GPT-6 LunaMuse Spark 1.3 Contributor
025507510050100150Perfect run6166

Every test case

153 test cases

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

15 won by GPT-6 Luna25 won by Muse Spark 1.3 Contributor113 tied