DeepSeek V4.1 Flash 1017 vs MiMo-V2.6-Pro, head to head

Side A

Side B

Overall

DeepSeek V4.1 Flash 1017 is clearly better. It wins 34 of the 53 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

TaskADeepSeek V4.1 Flash 1017BMiMo-V2.6-ProDeepSeek V4.1 Flash 1017MiMo-V2.6-Pro

Overall

5626.0s

5.3K5.9K

482m 13s

Agents

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

5826.7s

3.3K8.7K

474m 48s

Structured Extraction

Turns logs, diffs and records into exact JSON

8117.2s

4.2K3.4K

601m 26s

Coding

Writes code and SQL that pass hidden tests

4840.8s

9.4K10.7K

242m 58s

Grounded Answers

Answers from the given documents with sources, never guesses

7119.8s

4.4K3.2K

621m 9s

Instruction Following

Applies every rule, including ones changed mid-conversation

7625.6s

5.5K4.5K

591m 23s

Business Logic

Computes billing, scheduling and pricing rules exactly

6726.6s

6.6K7.1K

571m 52s

Security & permissions

Ignores planted instructions, still does the allowed work

4826.6s

4.7K3.6K

711m 41s

Diagnosis & investigation

Chooses diagnostic checks, repairs supported causes and verifies recovery

020.2s

3.2K6.2K

02m 54s

What you'd run

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

SpecADeepSeek V4.1 Flash 1017BMiMo-V2.6-Pro

Speed

243 tok/s

55 tok/s

Machine

Hosted

Memory at 32K

—

Context

1.1M

Cost per run

$0.75

$1.80

Cost per case

$0.0049

$0.012

Time per case

26.0s

2m 13s

Quantization

—

BF16

Harness

openrouter

The race

1 lead change

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

DeepSeek V4.1 Flash 1017MiMo-V2.6-Pro

Every test case

153 test cases

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

Agents52

Structured Extraction94

Coding61

Grounded Answers42

Instruction Following53

Business Logic31

Security & permissions26

Diagnosis & investigation00

34 won by DeepSeek V4.1 Flash 101719 won by MiMo-V2.6-Pro100 tied