[mle]/DaBench

Clover 35B A3B Preview vs Mistral Large 4

Side A
Model page
Side B
Model page
Overall
30
vs
Overall
28

Too close to call. Each wins 26 of the 52 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. The time under each score is the average per test case, on the machine it ran on. Clover 35B A3B Preview spends 1.8× as many thinking tokens per test case.

Thinking tokensAnswer tokens
TaskAB
Overall
3035.1s
8.4K, 95% thinking6.2K, 72% thinking
281m 56s
Agents
2036.8s
9.6K, 91% thinking795, no thinking
1126.1s
Structured Extraction
4327.5s
7.2K, 97% thinking288, no thinking
153.78s
Coding
101m 5s
12.2K, 100% thinking9.3K, 37% thinking
332m 48s
Grounded Answers
4821.0s
5.4K, 94% thinking8.1K, 89% thinking
382m 45s
Instruction Following
5332.8s
9.2K, 95% thinking12.9K, 84% thinking
693m 47s
Business Logic
1936.1s
9.8K, 99% thinking8.8K, 77% thinking
242m 40s
Security & permissions
4725.6s
5.8K, 92% thinking3.7K, 83% thinking
331m 9s
Diagnosis & investigation
037.0s
7K, 82% thinking4.4K, 85% thinking
01m 19s

What you'd run

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

SpecAB
Speed
252 tok/s (better)
56 tok/s
Machine
Hosted
Hosted
Memory at 32K
—
—
Context
—
512K
Cost per run
Not measured
$2.67
Cost per case
Not measured
$0.017
Time per case
35.1s (better)
1m 56s
Quantization
—
—
Harness
custom
openrouter

The race

No lead changes

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

Clover 35B A3B PreviewMistral Large 4
0255075100501001503028

Every test case

153 test cases

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

26 won by Clover 35B A3B Preview26 won by Mistral Large 4101 tied