[mle]/DaBench

Clover 35B A3B Preview vs GLM 5.3 Flash

Side A
Model page
Side B
Model page
Overall
30
vs
Overall
35

Too close to call. GLM 5.3 Flash wins 32 of the 53 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 3.2× as many thinking tokens per test case.

Thinking tokensAnswer tokens
TaskAB
Overall
3035.1s
8.4K, 95% thinking2.8K, 89% thinking
3555.5s
Agents
2036.8s
9.6K, 91% thinking607, 81% thinking
1037.0s
Structured Extraction
4327.5s
7.2K, 97% thinking2.7K, 93% thinking
4847.2s
Coding
101m 5s
12.2K, 100% thinking5.3K, 83% thinking
541m 52s
Grounded Answers
4821.0s
5.4K, 94% thinking2.9K, 93% thinking
3756.8s
Instruction Following
5332.8s
9.2K, 95% thinking2K, 77% thinking
4340.8s
Business Logic
1936.1s
9.8K, 99% thinking5.8K, 98% thinking
331m 18s
Security & permissions
4725.6s
5.8K, 92% thinking963, 81% thinking
5525.2s
Diagnosis & investigation
037.0s
7K, 82% thinking563, 63% thinking
027.0s

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)
60 tok/s
Machine
Hosted
Hosted
Memory at 32K
—
—
Context
—
1M
Cost per run
Not measured
$0.19
Cost per case
Not measured
$0.0013
Time per case
35.1s (better)
55.5s
Quantization
—
FP8
Harness
custom
openrouter

The race

2 lead changes

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

Clover 35B A3B PreviewGLM 5.3 Flash
0255075100501001503035

Every test case

153 test cases

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

21 won by Clover 35B A3B Preview32 won by GLM 5.3 Flash100 tied