[mle]/DaBench

DeepSeek V4.1 Flash 1017 vs Muse Spark 1.3 Contributor

Side A
Model page
Side B
Model page
Overall
56
vs
Overall
66

Muse Spark 1.3 Contributor is clearly better. It wins 29 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 1.6× as many thinking tokens per test case.

Thinking tokensAnswer tokens
TaskAB
Overall
5626.0s
5.3K, 91% thinking8.7K, 91% thinking
661m 42s
Agents
5826.7s
3.3K, 71% thinking14.7K, 88% thinking
573m 16s
Structured Extraction
8117.2s
4.2K, 93% thinking4.2K, 91% thinking
9037.9s
Coding
4840.8s
9.4K, 95% thinking10.8K, 95% thinking
482m 1s
Grounded Answers
7119.8s
4.4K, 91% thinking6.1K, 91% thinking
811m 1s
Instruction Following
7625.6s
5.5K, 91% thinking8.7K, 91% thinking
841m 18s
Business Logic
6726.6s
6.6K, 97% thinking6K, 95% thinking
761m 7s
Security & permissions
4826.6s
4.7K, 87% thinking8.2K, 87% thinking
761m 59s
Diagnosis & investigation
020.2s
3.2K, 74% 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
243 tok/s (better)
107 tok/s
Machine
Hosted
Hosted
Memory at 32K
—
—
Context
1M
1M
Cost per run
$0.75
$0.41 (better)
Cost per case
$0.0049
$0.0027 (better)
Time per case
26.0s (better)
1m 42s
Quantization
—
—
Harness
openrouter
openrouter

The race

4 lead changes

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

DeepSeek V4.1 Flash 1017Muse Spark 1.3 Contributor
025507510050100150Perfect run5666

Every test case

153 test cases

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

11 won by DeepSeek V4.1 Flash 101729 won by Muse Spark 1.3 Contributor113 tied