Mellum2.1 Thinking vs Spark X2.6 4B, head to head

Side A

Side B

Overall

Spark X2.6 4B is clearly better. It wins 23 of the 33 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. Both models spend about the same on thinking per test case.

Thinking tokensAnswer tokens

TaskAMellum2.1 ThinkingBSpark X2.6 4BMellum2.1 ThinkingSpark X2.6 4B

Overall

91m 56s

10.3K8.2K

151m 3s

Agents

Runs multi-step jobs through tools, within a written policy

92m 3s

10.8K11.5K

141m 32s

Structured Extraction

Turns logs, diffs and records into exact JSON

51m 36s

8.3K5.3K

1039.5s

Coding

Writes code and SQL that pass hidden tests

52m 11s

11.3K9.1K

41m 8s

Grounded Answers

Answers from the given documents with sources, never guesses

131m 24s

7.3K6.6K

1050.2s

Instruction Following

Applies every rule, including ones changed mid-conversation

283m 10s

17.3K11.9K

371m 29s

Business Logic

Computes billing, scheduling and pricing rules exactly

01m 39s

9K4.9K

1036.3s

Security & permissions

Ignores planted instructions, still does the allowed work

81m 40s

9.1K7K

3154.3s

Diagnosis & investigation

Chooses diagnostic checks, repairs supported causes and verifies recovery

01m 15s

6.8K11.6K

01m 30s

What you'd run

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

SpecAMellum2.1 ThinkingBSpark X2.6 4B

Speed

90 tok/s

136 tok/s

Machine

Apple M5 Pro, 48 GB

RTX 5090, 32 GB

Memory at 32K

12.9 GB

9.4 GB

Context

128K

605K

Cost per run

$2.46 est.

$1.61 est.

Cost per case

$0.016 est.

$0.011 est.

Time per case

1m 56s

1m 3s

Quantization

Q8_0

BF16

Harness

llama.cpp

The race

4 lead changes

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

Mellum2.1 ThinkingSpark X2.6 4B

Every test case

153 test cases

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

Agents13

Structured Extraction12

Coding23

Grounded Answers22

Instruction Following24

Business Logic02

Security & permissions27

Diagnosis & investigation00

10 won by Mellum2.1 Thinking23 won by Spark X2.6 4B120 tied