[mle]/DaBench

Qwen3.8 Flash Next vs Claude Haiku 5.5

Side A
Model page
Side B
Model page
Overall
42
vs
Overall
70

Claude Haiku 5.5 is clearly better. It wins 69 of the 78 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. The time under each score is the average per test case, on the machine it ran on. Both models spend about the same on thinking per test case.

Thinking tokensAnswer tokens
TaskAB
Overall
421m 15s
4.1K, 80% thinking3.9K, 78% thinking
7025.5s
Agents
522m 4s
4.6K, 78% thinking3.7K, 61% thinking
6139.5s
Structured Extraction
5042.2s
3.1K, 90% thinking2.6K, 83% thinking
9013.5s
Coding
521m 36s
6.2K, 61% thinking6.8K, 76% thinking
8630.6s
Grounded Answers
5147.0s
3.1K, 88% thinking2.2K, 78% thinking
7613.9s
Instruction Following
621m 30s
5.2K, 88% thinking4.3K, 82% thinking
9431.5s
Business Logic
251m 0s
3.9K, 90% thinking5.5K, 95% thinking
6726.0s
Security & permissions
451m 8s
2.8K, 82% thinking2.7K, 70% thinking
6723.4s
Diagnosis & investigation
01m 15s
3.4K, 69% 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
68 tok/s
225 tok/s (better)
Machine
Hosted
Hosted
Memory at 32K
—
—
Context
1M
1M
Cost per run
$0.46
$0.42 (better)
Cost per case
$0.003
$0.0028 (better)
Time per case
1m 15s
25.5s (better)
Quantization
—
—
Harness
openrouter
openrouter

The race

2 lead changes

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

Qwen3.8 Flash NextClaude Haiku 5.5
0255075100501001504270

Every test case

153 test cases

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

9 won by Qwen3.8 Flash Next69 won by Claude Haiku 5.575 tied