[mle]/DaBench

Mistral Large 4 vs Spark X2.6 4B

Side A
Model page
Side B
Model page
Overall
28
vs
Overall
15

Mistral Large 4 is clearly better. It wins 38 of the 48 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 1.5× as many thinking tokens per test case.

Thinking tokensAnswer tokens
TaskAB
Overall
281m 56s
6.2K, 72% thinking8.2K, 84% thinking
151m 3s
Agents
1126.1s
795, no thinking11.5K, 97% thinking
141m 32s
Structured Extraction
153.78s
288, no thinking5.3K, 78% thinking
1039.5s
Coding
332m 48s
9.3K, 37% thinking9.1K, 48% thinking
41m 8s
Grounded Answers
382m 45s
8.1K, 89% thinking6.6K, 86% thinking
1050.2s
Instruction Following
693m 47s
12.9K, 84% thinking11.9K, 90% thinking
371m 29s
Business Logic
242m 40s
8.8K, 77% thinking4.9K, 86% thinking
1036.3s
Security & permissions
331m 9s
3.7K, 83% thinking7K, 94% thinking
3154.3s
Diagnosis & investigation
01m 19s
4.4K, 85% thinking11.6K, 98% thinking
01m 30s

What you'd run

The build behind each score. Orange marks the faster, smaller, cheaper or longer-context side.

SpecAB
Speed
56 tok/s
136 tok/s
Machine
Hosted
RTX 5090, 32 GB
Memory at 32K
—
9.4 GB
Context
512K
605K (better)
Cost per run
$2.67
$1.61 est. (better)
Cost per case
$0.017
$0.011 est. (better)
Time per case
1m 56s
1m 3s (better)
Quantization
—
BF16
Harness
openrouter
llama.cpp

The race

3 lead changes

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

Mistral Large 4Spark X2.6 4B
0255075100501001502815

Every test case

153 test cases

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

38 won by Mistral Large 410 won by Spark X2.6 4B105 tied