Qwen3.8 27B vs Underdog Saluki 27B, head to head

Side A

Side B

Overall

Qwen3.8 27B is clearly better. It wins 44 of the 67 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

TaskAQwen3.8 27BBUnderdog Saluki 27BQwen3.8 27BUnderdog Saluki 27B

Overall

4833.3s

5.9K6.9K

335m 15s

Agents

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

5736.6s

5.2K11.8K

409m 2s

Structured Extraction

Turns logs, diffs and records into exact JSON

7320.6s

4.6K4.4K

433m 17s

Coding

Writes code and SQL that pass hidden tests

3753.0s

9.3K9.2K

166m 54s

Grounded Answers

Answers from the given documents with sources, never guesses

6131.6s

4.2K4.5K

433m 22s

Instruction Following

Applies every rule, including ones changed mid-conversation

6733.6s

6.9K8K

466m 5s

Business Logic

Computes billing, scheduling and pricing rules exactly

3632.9s

7.6K5.2K

333m 55s

Security & permissions

Ignores planted instructions, still does the allowed work

5122.6s

3.7K4.9K

433m 47s

Diagnosis & investigation

Chooses diagnostic checks, repairs supported causes and verifies recovery

040.0s

5.9K8.7K

06m 39s

What you'd run

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

SpecAQwen3.8 27BBUnderdog Saluki 27B

Speed

196 tok/s

23 tok/s

Machine

Hosted

Apple M5 Pro, 48 GB

Memory at 32K

—

10.2 GB

Context

256K

Cost per run

$1.28

Not measured

Cost per case

$0.0083

Not measured

Time per case

33.3s

5m 15s

Quantization

FP8

IQ2_XS

Harness

openrouter

llama.cpp

The race

1 lead change

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

Qwen3.8 27BUnderdog Saluki 27B

Every test case

153 test cases

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

Agents71

Structured Extraction105

Coding84

Grounded Answers51

Instruction Following76

Business Logic45

Security & permissions31

Diagnosis & investigation00

44 won by Qwen3.8 27B23 won by Underdog Saluki 27B86 tied