[mle]/DaBench

GLM 5.3 Flash vs Qwen3.8 Flash

Side A
Model page
Side B
Model page
Overall
35
vs
Overall
32

Too close to call. GLM 5.3 Flash wins 40 of the 72 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. Qwen3.8 Flash spends 1.3× as many thinking tokens per test case.

Thinking tokensAnswer tokens
TaskAB
Overall
3555.5s
2.8K, 89% thinking4K, 81% thinking
3258.5s
Agents
1037.0s
607, 81% thinking4.2K, 76% thinking
491m 25s
Structured Extraction
4847.2s
2.7K, 93% thinking3K, 90% thinking
3336.9s
Coding
541m 52s
5.3K, 83% thinking5.8K, 61% thinking
471m 21s
Grounded Answers
3756.8s
2.9K, 93% thinking3.1K, 90% thinking
2641.4s
Instruction Following
4340.8s
2K, 77% thinking5K, 89% thinking
371m 4s
Business Logic
331m 18s
5.8K, 98% thinking3.9K, 91% thinking
1746.8s
Security & permissions
5525.2s
963, 81% thinking3.2K, 83% thinking
5054.1s
Diagnosis & investigation
027.0s
563, 63% 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
60 tok/s
77 tok/s (better)
Machine
Hosted
Hosted
Memory at 32K
—
—
Context
1M
1M
Cost per run
$0.19 (better)
$0.46
Cost per case
$0.0013 (better)
$0.003
Time per case
55.5s (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.

GLM 5.3 FlashQwen3.8 Flash
025507510050100150Perfect run3532

Every test case

153 test cases

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

40 won by GLM 5.3 Flash32 won by Qwen3.8 Flash81 tied