Model cards · 17 setups

Every model, its own page.

Pick one. See what it scored, where it slipped, and how fast it runs on the test bench.

  1. 01Current teamgpt-oss-120bMXFP448.5 tok/s98.4
  2. 02Current teamGemma 4 31B QATUD-Q4_K_XL10.6 tok/s94.4
  3. 03ContenderQwen3-Coder-NextUD-Q4_K_XL46.8 tok/s91.1
  4. 04ContenderGLM-5.3-FlashUD-IQ3_XXS13.4 tok/s89.2
  5. 05RetiredKAT-Coder v2.5Q8_051.4 tok/s83.0
  6. 06RetiredQwen3-Coder-NextQ641 tok/s83.0
  7. 07ContenderMuse Glimmer 30BQ8_07.6 tok/s82.7
  8. 08RetiredTiel-Coder 35BQ8_043.6 tok/s81.4
  9. 09ContenderNemotron 3 Super 120BUD-Q4_K_XL16.8 tok/s81.0
  10. 10ContenderDeepSeek V4-FlashUD-IQ3_XXS11.8 tok/s79.7
  11. 11RetiredQwen3.8-Flash-NextIQ4_XS-PLE · fixed sampling28.2 tok/s78.4
  12. 12RetiredQwen3.8-Flash-NextUD-IQ4_XS · fixed sampling23.6 tok/s77.7
  13. 13RetiredGemma 4 26B-A4B QATUD-Q4_K_XL57.5 tok/s71.0
  14. 14RetiredDevstral 2 24BQ8_08.5 tok/s70.0
  15. 15Under auditQwen3.8-27BQ8_07.5 tok/s66.0
  16. 16RetiredGemma 4 12B QATUD-Q4_K_XL25.5 tok/s62.0
  17. 17RetiredMistral Small 4 119BUD-Q4_K_XL36.1 tok/s61.7