Head to head

Any two models from the board, on one scale.

Model A

Model B

openaihosted

overall

12 ▸

openrouter · deepseekhosted

48vs60

DeepSeek V4.1 Flash leads by 12 overall · ahead on 9 of 9 tasks

the tape

Every case, one strip

2 cases

Qwen3.8 27B takes 0DeepSeek V4.1 Flash takes 00 even2 ran on one side only

one cell per case, in suite order · the brighter score wins the cell · hover for the case

tasks

9 tasks

65TOOL+13 ▸78

48STRUCT+4 ▸52

59RAG+4 ▸63

94CTX+6 ▸100

7CODE+34 ▸41

32REASON+7 ▸39

25INSTR+21 ▸46

38CLASS+17 ▸55

62SUM+8 ▸70

best build per side · same suite, same graders · hover a score for cases passed

tokens

Qwen3.8 27B

0 tok/case · no visible thinking

DeepSeek V4.1 Flash

not run

0REASONnot run—

Neither model shows visible thinking on this suite

bar = average output tokens per case, shared scale · solid = answer · faded = thinking · hover a bar for the split

specs

45 tok/sspeed215 tok/s

—memory—

—context1.0M

$0cost / run$0.7017

—quantization—

openaiharnessopenrouter

the lit value is the better spec — faster, smaller, cheaper, or more context · quantization & harness just describe the builds