[mle]/DaBench

GLM 5.3 Flash vs Mistral Large 4

Side A
Model page
Side B
Model page
Overall
35
vs
Overall
28

Too close to call. GLM 5.3 Flash wins 37 of the 65 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. Mistral Large 4 spends 1.8× as many thinking tokens per test case.

Thinking tokensAnswer tokens
TaskAB
Overall
3555.5s
2.8K, 89% thinking6.2K, 72% thinking
281m 56s
Agents
1037.0s
607, 81% thinking795, no thinking
1126.1s
Structured Extraction
4847.2s
2.7K, 93% thinking288, no thinking
153.78s
Coding
541m 52s
5.3K, 83% thinking9.3K, 37% thinking
332m 48s
Grounded Answers
3756.8s
2.9K, 93% thinking8.1K, 89% thinking
382m 45s
Instruction Following
4340.8s
2K, 77% thinking12.9K, 84% thinking
693m 47s
Business Logic
331m 18s
5.8K, 98% thinking8.8K, 77% thinking
242m 40s
Security & permissions
5525.2s
963, 81% thinking3.7K, 83% thinking
331m 9s
Diagnosis & investigation
027.0s
563, 63% 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
60 tok/s (better)
56 tok/s
Machine
Hosted
Hosted
Memory at 32K
—
—
Context
1M (better)
512K
Cost per run
$0.19 (better)
$2.67
Cost per case
$0.0013 (better)
$0.017
Time per case
55.5s (better)
1m 56s
Quantization
FP8
—
Harness
openrouter
openrouter

The race

2 lead changes

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

GLM 5.3 FlashMistral Large 4
025507510050100150Perfect run3528

Every test case

153 test cases

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

37 won by GLM 5.3 Flash28 won by Mistral Large 488 tied