[mle]/DaBench

GLM 5.3 Flash vs GPT-6 Luna

Side A
Model page
Side B
Model page
Overall
35
vs
Overall
61

GPT-6 Luna is clearly better. It wins 61 of the 69 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. GLM 5.3 Flash spends 1.8× as many thinking tokens per test case.

Thinking tokensAnswer tokens
TaskAB
Overall
3555.5s
2.8K, 89% thinking1.9K, 73% thinking
6149.1s
Agents
1037.0s
607, 81% thinking1.3K, 67% thinking
561m 18s
Structured Extraction
4847.2s
2.7K, 93% thinking999, 72% thinking
9514.5s
Coding
541m 52s
5.3K, 83% thinking3.6K, 60% thinking
732m 18s
Grounded Answers
3756.8s
2.9K, 93% thinking1.1K, 72% thinking
6715.6s
Instruction Following
4340.8s
2K, 77% thinking2K, 74% thinking
8432.3s
Business Logic
331m 18s
5.8K, 98% thinking3.2K, 93% thinking
3743.1s
Security & permissions
5525.2s
963, 81% thinking1K, 68% thinking
6224.5s
Diagnosis & investigation
027.0s
563, 63% thinking1.1K, 68% thinking
1737.7s

What you'd run

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

SpecAB
Speed
60 tok/s
90 tok/s (better)
Machine
Hosted
Hosted
Memory at 32K
—
—
Context
1M
1.1M (better)
Cost per run
$0.19
$0.10 (better)
Cost per case
$0.0013
$0.00067 (better)
Time per case
55.5s
49.1s (better)
Quantization
FP8
—
Harness
openrouter
openrouter

The race

No lead changes

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

GLM 5.3 FlashGPT-6 Luna
025507510050100150Perfect run3561

Every test case

153 test cases

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

8 won by GLM 5.3 Flash61 won by GPT-6 Luna84 tied