MiMo-V2.6-Flash vs Spark X2.6 4B, head to head

Side A

Side B

Overall

MiMo-V2.6-Flash is clearly better. It wins 62 of the 68 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 1.8× as many thinking tokens per test case.

Thinking tokensAnswer tokens

TaskAMiMo-V2.6-FlashBSpark X2.6 4BMiMo-V2.6-FlashSpark X2.6 4B

Overall

451m 21s

4.2K8.2K

151m 3s

Agents

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

381m 19s

2.2K11.5K

141m 32s

Structured Extraction

Turns logs, diffs and records into exact JSON

6238.0s

2.6K5.3K

1039.5s

Coding

Writes code and SQL that pass hidden tests

242m 40s

9.9K9.1K

41m 8s

Grounded Answers

Answers from the given documents with sources, never guesses

6242.1s

2.6K6.6K

1050.2s

Instruction Following

Applies every rule, including ones changed mid-conversation

641m 2s

4.8K11.9K

371m 29s

Business Logic

Computes billing, scheduling and pricing rules exactly

4356.4s

4.4K4.9K

1036.3s

Security & permissions

Ignores planted instructions, still does the allowed work

662m 6s

3.8K7K

3154.3s

Diagnosis & investigation

Chooses diagnostic checks, repairs supported causes and verifies recovery

01m 33s

1.5K11.6K

01m 30s

What you'd run

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

SpecAMiMo-V2.6-FlashBSpark X2.6 4B

Speed

66 tok/s

136 tok/s

Machine

Hosted

RTX 5090, 32 GB

Memory at 32K

—

9.4 GB

Context

1.1M

605K

Cost per run

$0.33

$1.61 est.

Cost per case

$0.0021

$0.011 est.

Time per case

1m 21s

1m 3s

Quantization

BF16

Harness

openrouter

llama.cpp

The race

No lead changes

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

MiMo-V2.6-FlashSpark X2.6 4B

Every test case

153 test cases

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

Agents61

Structured Extraction121

Coding51

Grounded Answers110

Instruction Following101

Business Logic81

Security & permissions101

Diagnosis & investigation00

62 won by MiMo-V2.6-Flash6 won by Spark X2.6 4B85 tied