[mle]/DaBench

Beam 1.0 501B A23B vs Qwen3.8 Flash

Side A
Model page
Side B
Model page
Overall
57
vs
Overall
32

Beam 1.0 501B A23B is clearly better. It wins 68 of the 88 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. Both models spend about the same on thinking per test case.

Thinking tokensAnswer tokens
TaskAB
Overall
5717.2s
4K, 87% thinking4K, 81% thinking
3258.5s
Agents
5521.8s
4.8K, 90% thinking4.2K, 76% thinking
491m 25s
Structured Extraction
819.98s
3.4K, 91% thinking3K, 90% thinking
3336.9s
Coding
4819.9s
4.5K, 70% thinking5.8K, 61% thinking
471m 21s
Grounded Answers
6911.6s
2.7K, 88% thinking3.1K, 90% thinking
2641.4s
Instruction Following
7820.9s
5.6K, 90% thinking5K, 89% thinking
371m 4s
Business Logic
5714.0s
4.2K, 95% thinking3.9K, 91% thinking
1746.8s
Security & permissions
5120.6s
3.2K, 88% thinking3.2K, 83% thinking
5054.1s
Diagnosis & investigation
1724.3s
3.1K, 90% thinking3.3K, 73% thinking
058.4s

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)
77 tok/s
Machine
Hosted
Hosted
Memory at 32K
—
—
Context
1M
1M
Cost per run
$0.83
$0.46 (better)
Cost per case
$0.0054
$0.003 (better)
Time per case
17.2s (better)
58.5s
Quantization
FP8
—
Harness
openrouter
openrouter

The race

3 lead changes

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

Beam 1.0 501B A23BQwen3.8 Flash
025507510050100150Perfect run5732

Every test case

153 test cases

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

68 won by Beam 1.0 501B A23B20 won by Qwen3.8 Flash65 tied