[mle]/DaBench

Claude Haiku 5.5 vs Mistral Large 4

Side A
Model page
Side B
Model page
Overall
70
vs
Overall
28

Claude Haiku 5.5 is clearly better. It wins 74 of the 76 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. Mistral Large 4 spends 1.5× as many thinking tokens per test case.

Thinking tokensAnswer tokens
TaskAB
Overall
7025.5s
3.9K, 78% thinking6.2K, 72% thinking
281m 56s
Agents
6139.5s
3.7K, 61% thinking795, no thinking
1126.1s
Structured Extraction
9013.5s
2.6K, 83% thinking288, no thinking
153.78s
Coding
8630.6s
6.8K, 76% thinking9.3K, 37% thinking
332m 48s
Grounded Answers
7613.9s
2.2K, 78% thinking8.1K, 89% thinking
382m 45s
Instruction Following
9431.5s
4.3K, 82% thinking12.9K, 84% thinking
693m 47s
Business Logic
6726.0s
5.5K, 95% thinking8.8K, 77% thinking
242m 40s
Security & permissions
6723.4s
2.7K, 70% thinking3.7K, 83% thinking
331m 9s
Diagnosis & investigation
1725.5s
2.3K, 50% thinking4.4K, 85% thinking
01m 19s

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)
56 tok/s
Machine
Hosted
Hosted
Memory at 32K
—
—
Context
1M (better)
512K
Cost per run
$0.42 (better)
$2.67
Cost per case
$0.0028 (better)
$0.017
Time per case
25.5s (better)
1m 56s
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.

Claude Haiku 5.5Mistral Large 4
025507510050100150Perfect run7028

Every test case

153 test cases

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

74 won by Claude Haiku 5.52 won by Mistral Large 477 tied