[mle]/DaBench

Ling 3.1 Flash vs MiniMax M3.1 Flash

Side A
Model page
Side B
Model page
Overall
45
vs
Overall
35

Ling 3.1 Flash is clearly better. It wins 34 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. Under each score, the average time per test case on the machine it ran on. Ling 3.1 Flash spends 3.4× as many thinking tokens per test case.

Thinking tokensAnswer tokens
TaskAB
Overall
451m 16s
7.9K, 73% thinking2.3K, 73% thinking
3521.3s
Agents
4155.9s
4.5K, 66% thinking3.1K, 71% thinking
3637.0s
Structured Extraction
711m 11s
8.1K, 69% thinking1.1K, 74% thinking
487.96s
Coding
241m 42s
11.3K, 89% thinking4.6K, 66% thinking
5840.6s
Grounded Answers
5258.9s
6.7K, 74% thinking1.2K, 71% thinking
339.75s
Instruction Following
761m 55s
11.4K, 77% thinking1.7K, 70% thinking
5314.7s
Business Logic
381m 22s
9.7K, 56% thinking3.5K, 94% thinking
2425.9s
Security & permissions
5456.1s
5.2K, 74% thinking1.3K, 60% thinking
3012.7s
Diagnosis & investigation
042.1s
3.2K, 72% thinking2.1K, 51% thinking
022.8s

What you'd run

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

SpecAB
Speed
112 tok/s
135 tok/s (better)
Machine
Hosted
Hosted
Memory at 32K
—
—
Context
256K
1M (better)
Cost per run
Not measured
Not measured
Cost per case
Not measured
Not measured
Time per case
1m 16s
21.3s (better)
Quantization
—
—
Harness
openrouter
openrouter

The race

5 lead changes

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

Ling 3.1 FlashMiniMax M3.1 Flash
025507510050100150Perfect run4535

Every test case

153 test cases

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

34 won by Ling 3.1 Flash19 won by MiniMax M3.1 Flash100 tied