[mle]/DaBench

Claude Haiku 5.5 vs MiniMax M3.1 Flash

Side A
Model page
Side B
Model page
Overall
70
vs
Overall
35

Claude Haiku 5.5 is clearly better. It wins 68 of the 75 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. Claude Haiku 5.5 spends 1.8× as many thinking tokens per test case.

Thinking tokensAnswer tokens
TaskAB
Overall
7025.5s
3.9K, 78% thinking2.3K, 73% thinking
3521.3s
Agents
6139.5s
3.7K, 61% thinking3.1K, 71% thinking
3637.0s
Structured Extraction
9013.5s
2.6K, 83% thinking1.1K, 74% thinking
487.96s
Coding
8630.6s
6.8K, 76% thinking4.6K, 66% thinking
5840.6s
Grounded Answers
7613.9s
2.2K, 78% thinking1.2K, 71% thinking
339.75s
Instruction Following
9431.5s
4.3K, 82% thinking1.7K, 70% thinking
5314.7s
Business Logic
6726.0s
5.5K, 95% thinking3.5K, 94% thinking
2425.9s
Security & permissions
6723.4s
2.7K, 70% thinking1.3K, 60% thinking
3012.7s
Diagnosis & investigation
1725.5s
2.3K, 50% 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
225 tok/s (better)
135 tok/s
Machine
Hosted
Hosted
Memory at 32K
—
—
Context
1M
1M
Cost per run
$0.42
Not measured
Cost per case
$0.0028
Not measured
Time per case
25.5s
21.3s (better)
Quantization
—
—
Harness
openrouter
openrouter

The race

1 lead change

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

Claude Haiku 5.5MiniMax M3.1 Flash
025507510050100150Perfect run7035

Every test case

153 test cases

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

68 won by Claude Haiku 5.57 won by MiniMax M3.1 Flash78 tied