DeepSeek V4.1 Flash vs Mellum2.1 Thinking, head to head

Side A

Side B

Overall

DeepSeek V4.1 Flash is clearly better. It wins 98 of the 99 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 1.2× as many thinking tokens per test case.

Thinking tokensAnswer tokens

TaskADeepSeek V4.1 FlashBMellum2.1 ThinkingDeepSeek V4.1 FlashMellum2.1 Thinking

Overall

643m 35s

6K10.3K

91m 56s

Agents

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

464m 42s

6.1K10.8K

92m 3s

Structured Extraction

Turns logs, diffs and records into exact JSON

862m 8s

4.4K8.3K

51m 36s

Coding

Writes code and SQL that pass hidden tests

514m 55s

9.7K11.3K

52m 11s

Grounded Answers

Answers from the given documents with sources, never guesses

762m 17s

3.9K7.3K

131m 24s

Instruction Following

Applies every rule, including ones changed mid-conversation

894m 24s

7.5K17.3K

283m 10s

Business Logic

Computes billing, scheduling and pricing rules exactly

814m 1s

7K9K

01m 39s

Security & permissions

Ignores planted instructions, still does the allowed work

693m 3s

4.4K9.1K

81m 40s

Diagnosis & investigation

Chooses diagnostic checks, repairs supported causes and verifies recovery

172m 8s

3.2K6.8K

01m 15s

What you'd run

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

SpecADeepSeek V4.1 FlashBMellum2.1 Thinking

Speed

30 tok/s

90 tok/s

Machine

Hosted

Apple M5 Pro, 48 GB

Memory at 32K

—

12.9 GB

Context

128K

Cost per run

$0.39

$2.46 est.

Cost per case

$0.0026

$0.016 est.

Time per case

3m 35s

1m 56s

Quantization

FP8

Q8_0

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.

DeepSeek V4.1 FlashMellum2.1 Thinking

Every test case

153 test cases

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

Agents90

Structured Extraction170

Coding101

Grounded Answers140

Instruction Following150

Business Logic170

Security & permissions150

Diagnosis & investigation10

98 won by DeepSeek V4.1 Flash1 won by Mellum2.1 Thinking54 tied