[mle]/DaBench

Spark X2.6 4B vs Claude Haiku 5.5

Side A
Model page
Side B
Model page
Overall
15
vs
Overall
70

Claude Haiku 5.5 is clearly better. It wins 99 of the 101 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. Spark X2.6 4B spends 2.3× as many thinking tokens per test case.

Thinking tokensAnswer tokens
TaskAB
Overall
151m 3s
8.2K, 84% thinking3.9K, 78% thinking
7025.5s
Agents
141m 32s
11.5K, 97% thinking3.7K, 61% thinking
6139.5s
Structured Extraction
1039.5s
5.3K, 78% thinking2.6K, 83% thinking
9013.5s
Coding
41m 8s
9.1K, 48% thinking6.8K, 76% thinking
8630.6s
Grounded Answers
1050.2s
6.6K, 86% thinking2.2K, 78% thinking
7613.9s
Instruction Following
371m 29s
11.9K, 90% thinking4.3K, 82% thinking
9431.5s
Business Logic
1036.3s
4.9K, 86% thinking5.5K, 95% thinking
6726.0s
Security & permissions
3154.3s
7K, 94% thinking2.7K, 70% thinking
6723.4s
Diagnosis & investigation
01m 30s
11.6K, 98% 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
136 tok/s
225 tok/s
Machine
RTX 5090, 32 GB
Hosted
Memory at 32K
9.4 GB
—
Context
605K
1M (better)
Cost per run
$1.61 est.
$0.42 (better)
Cost per case
$0.011 est.
$0.0028 (better)
Time per case
1m 3s
25.5s (better)
Quantization
BF16
—
Harness
llama.cpp
openrouter

The race

No lead changes

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

Spark X2.6 4BClaude Haiku 5.5
0255075100501001501570

Every test case

153 test cases

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

2 won by Spark X2.6 4B99 won by Claude Haiku 5.552 tied