PipelineScore
PipelineScore v3 · archive

24 models compared.

Best run per model. Click any column to re-rank, filter by provider, re-weight for your use case, and pick two rows to go head-to-head.

This is the 34-task v3 board, kept as an archive. The current test is v4. See the v4 board.

Weighting
24 models · sorted by Balanced composite (high → low) · pick any two rows to compare
#vsModelPipelineScore ▾Tier
1openai/gpt-oss-20blocal
93.2
TRUNK
2gpt-oss:latestlocal
91.6
TRUNK
3unsloth/gpt-oss-20b-GGUFlocal
91.3
TRUNK
4unsloth/gemma-4-E4B-it-GGUFlocal
89.5
DRIP
5Gemmalocal
88.4
MAINLINE
6qwen3.6-35b-a3blocal
87.5
MAINLINE
7qwen3-coder:30blocal
86.6
MAINLINE
8qwen3.8:27blocal
86.2
MAINLINE
9unsloth/Ornith-1.0-35B-GGUFlocal
86.1
MAINLINE
10unsloth/Ornith-1.0-9B-GGUFlocal
85.4
MAINLINE
11muse-glimmer:30blocal
84.9
MAINLINE
12unsloth/Qwen3.8-27B-GGUFlocal
84.7
MAINLINE
13qwen3.8-27b@iq3_xxslocal
84.2
MAINLINE
14mlx-community/Qwen3.6-35B-A3B-4bitlocal
83.7
MAINLINE
15unsloth/gemma-4-12b-it-GGUFlocal
83.2
MAINLINE
16/home/thomaskyn/models/Qwen3.6-35B-A3B/Qwen3.6-35B-A3B-UD-IQ3_S.gguflocal
80.9
MAINLINE
17gemma4:12b-it-qat_gpulocal
80.9
MAINLINE
18qwen/qwen3.6-35b-a3blocal
79.3
MAINLINE
19Jackrong/Qwen3.5-9B-DeepSeek-V4-Flash-GGUFlocal
79.3
MAINLINE
20unsloth/gemma-4-26B-A4B-it-GGUFlocal
78.8
MAINLINE
21qwen2.5-coder:7blocal
77.7
MAINLINE
22/home/thomaskyn/models/Qwen3.5-4b/Qwen3.5-4B-Q4_K_M.gguflocal
57.8
TAP
23/home/thomaskyn/models/Qwen3.5-9b/Qwen3.5-9B-Q4_K_M.gguflocal
50.5
TAP
24unsloth/North-Mini-Code-1.0-GGUFlocal
38.9
DRIP
vs