[mle]/DaBench

MiniMax M3.1 Flash vs Claude Haiku 5.5

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

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
3521.3s
2.3K, 73% thinking3.9K, 78% thinking
7025.5s
Agents
3637.0s
3.1K, 71% thinking3.7K, 61% thinking
6139.5s
Structured Extraction
487.96s
1.1K, 74% thinking2.6K, 83% thinking
9013.5s
Coding
5840.6s
4.6K, 66% thinking6.8K, 76% thinking
8630.6s
Grounded Answers
339.75s
1.2K, 71% thinking2.2K, 78% thinking
7613.9s
Instruction Following
5314.7s
1.7K, 70% thinking4.3K, 82% thinking
9431.5s
Business Logic
2425.9s
3.5K, 94% thinking5.5K, 95% thinking
6726.0s
Security & permissions
3012.7s
1.3K, 60% thinking2.7K, 70% thinking
6723.4s
Diagnosis & investigation
022.8s
2.1K, 51% 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
135 tok/s
225 tok/s (better)
Machine
Hosted
Hosted
Memory at 32K
—
—
Context
1M
1M
Cost per run
Not measured
$0.42
Cost per case
Not measured
$0.0028
Time per case
21.3s (better)
25.5s
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.

MiniMax M3.1 FlashClaude Haiku 5.5
025507510050100150Perfect run3570

Every test case

153 test cases

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

7 won by MiniMax M3.1 Flash68 won by Claude Haiku 5.578 tied