[mle]/DaBench

Ling 3.1 Flash vs Claude Haiku 5.5

Side A
Model page
Side B
Model page
Overall
45
vs
Overall
70

Claude Haiku 5.5 is clearly better. It wins 50 of the 56 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 1.9× as many thinking tokens per test case.

Thinking tokensAnswer tokens
TaskAB
Overall
451m 16s
7.9K, 73% thinking3.9K, 78% thinking
7025.5s
Agents
4155.9s
4.5K, 66% thinking3.7K, 61% thinking
6139.5s
Structured Extraction
711m 11s
8.1K, 69% thinking2.6K, 83% thinking
9013.5s
Coding
241m 42s
11.3K, 89% thinking6.8K, 76% thinking
8630.6s
Grounded Answers
5258.9s
6.7K, 74% thinking2.2K, 78% thinking
7613.9s
Instruction Following
761m 55s
11.4K, 77% thinking4.3K, 82% thinking
9431.5s
Business Logic
381m 22s
9.7K, 56% thinking5.5K, 95% thinking
6726.0s
Security & permissions
5456.1s
5.2K, 74% thinking2.7K, 70% thinking
6723.4s
Diagnosis & investigation
042.1s
3.2K, 72% thinking2.3K, 50% thinking
1725.5s

What you'd run

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

SpecAB
Speed
112 tok/s
225 tok/s (better)
Machine
Hosted
Hosted
Memory at 32K
—
—
Context
256K
1M (better)
Cost per run
Not measured
$0.42
Cost per case
Not measured
$0.0028
Time per case
1m 16s
25.5s (better)
Quantization
—
—
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.

Ling 3.1 FlashClaude Haiku 5.5
025507510050100150Perfect run4570

Every test case

153 test cases

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

6 won by Ling 3.1 Flash50 won by Claude Haiku 5.597 tied