[mle]/DaBench

Beam 1.0 501B A23B vs Muse Spark 1.3 Contributor

Side A
Model page
Side B
Model page
Overall
57
vs
Overall
66

Muse Spark 1.3 Contributor is clearly better. It wins 37 of the 48 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 2.2× as many thinking tokens per test case.

Thinking tokensAnswer tokens
TaskAB
Overall
5717.2s
4K, 87% thinking8.7K, 91% thinking
661m 42s
Agents
5521.8s
4.8K, 90% thinking14.7K, 88% thinking
573m 16s
Structured Extraction
819.98s
3.4K, 91% thinking4.2K, 91% thinking
9037.9s
Coding
4819.9s
4.5K, 70% thinking10.8K, 95% thinking
482m 1s
Grounded Answers
6911.6s
2.7K, 88% thinking6.1K, 91% thinking
811m 1s
Instruction Following
7820.9s
5.6K, 90% thinking8.7K, 91% thinking
841m 18s
Business Logic
5714.0s
4.2K, 95% thinking6K, 95% thinking
761m 7s
Security & permissions
5120.6s
3.2K, 88% thinking8.2K, 87% thinking
761m 59s
Diagnosis & investigation
1724.3s
3.1K, 90% 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
274 tok/s (better)
107 tok/s
Machine
Hosted
Hosted
Memory at 32K
—
—
Context
1M
1M
Cost per run
$0.83
$0.41 (better)
Cost per case
$0.0054
$0.0027 (better)
Time per case
17.2s (better)
1m 42s
Quantization
FP8
—
Harness
openrouter
openrouter

The race

6 lead changes

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

Beam 1.0 501B A23BMuse Spark 1.3 Contributor
025507510050100150Perfect run5766

Every test case

153 test cases

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

11 won by Beam 1.0 501B A23B37 won by Muse Spark 1.3 Contributor105 tied