Mellum2.1 Thinking vs MiMo-V2.6-Flash, head to head

Side A

Side B

Overall

MiMo-V2.6-Flash is clearly better. It wins 68 of the 71 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. Mellum2.1 Thinking spends 1.7× as many thinking tokens per test case.

Thinking tokensAnswer tokens

TaskAMellum2.1 ThinkingBMiMo-V2.6-FlashMellum2.1 ThinkingMiMo-V2.6-Flash

Overall

91m 56s

10.3K4.2K

451m 21s

Agents

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

92m 3s

10.8K2.2K

381m 19s

Structured Extraction

Turns logs, diffs and records into exact JSON

51m 36s

8.3K2.6K

6238.0s

Coding

Writes code and SQL that pass hidden tests

52m 11s

11.3K9.9K

242m 40s

Grounded Answers

Answers from the given documents with sources, never guesses

131m 24s

7.3K2.6K

6242.1s

Instruction Following

Applies every rule, including ones changed mid-conversation

283m 10s

17.3K4.8K

641m 2s

Business Logic

Computes billing, scheduling and pricing rules exactly

01m 39s

9K4.4K

4356.4s

Security & permissions

Ignores planted instructions, still does the allowed work

81m 40s

9.1K3.8K

662m 6s

Diagnosis & investigation

Chooses diagnostic checks, repairs supported causes and verifies recovery

01m 15s

6.8K1.5K

01m 33s

What you'd run

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

SpecAMellum2.1 ThinkingBMiMo-V2.6-Flash

Speed

90 tok/s

66 tok/s

Machine

Apple M5 Pro, 48 GB

Hosted

Memory at 32K

12.9 GB

—

Context

128K

1.1M

Cost per run

$2.46 est.

$0.33

Cost per case

$0.016 est.

$0.0021

Time per case

1m 56s

1m 21s

Quantization

Q8_0

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.

Mellum2.1 ThinkingMiMo-V2.6-Flash

Every test case

153 test cases

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

Agents18

Structured Extraction012

Coding14

Grounded Answers011

Instruction Following010

Business Logic09

Security & permissions114

Diagnosis & investigation00

3 won by Mellum2.1 Thinking68 won by MiMo-V2.6-Flash82 tied