MakoBench

Powered by CrucibleForge

Suite 3.0

MakoBench: Main Bench (suite 3.0)

Model Total Chat Code Params Date tested
deepseek-pro 90.2 87.6 93.4 API 2026-08-30
dark-scarlett-27b-v2 88.3 88.1 88.7 β€” 2026-08-24
minimax-m3 84.1 83.6 84.8 API 2026-08-30
qwen3.8-27b-abliterated 80.1 81.3 78.7 β€” 2026-08-22
split-untied-31b 80.0 88.7 69.4 31B 2026-09-09
dark-scarlett-iq4-xs-uncat 79.2 93.1 62.2 β€” 2026-09-04
joyfox-35b-rp 78.6 80.6 76.2 β€” 2026-08-23
jax-xortron-27b 78.5 90.1 64.4 β€” 2026-09-09
qwen3.8-27b-nvfp4-mtp 76.9 84.3 67.8 β€” 2026-09-08
qwen3.8-27b-unsloth 76.6 86.5 64.4 27B 2026-09-08
darker-scarlett-27b 76.5 86.5 64.4 27B 2026-09-08
gemma-e4b-uncensored 76.4 80.3 71.6 β€” 2026-08-22
orcarouter-qwen3.8-27b 75.9 87.0 62.2 β€” 2026-09-08
darker-hero-27b 75.1 85.6 62.2 27B 2026-09-08
dark-scarlett-35b-v2 74.4 – 74.4 27B 2026-09-09
deepseek-flash 72.2 85.9 55.6 API 2026-09-09
dark-scarlett-35b-iq4-xs-uncat 71.5 87.6 51.8 27B 2026-09-09
trohrbaugh-122b-heretic 70.6 87.1 25.0 122B-A10B 2026-09-04
muse-glimmer-30b 67.1 80.7 50.4 β€” 2026-08-22
orion-26b-a4b-v1.1 63.3 – 63.3 β€” 2026-09-09
laguna-s-vision 55.8 76.9 30.0 β€” 2026-09-09
cydonia-22b 44.8 74.2 8.9 β€” 2026-09-09
celeste-12b 39.6 68.3 4.4 β€” 2026-09-09
qwen2.5-vl-7b 39.6 55.5 20.3 β€” 2026-08-23
deepseek-r1-0528-8b 37.3 59.6 10.0 8B 2026-09-09
gemma-2-2b-abliterated 37.1 64.7 3.3 2B 2026-09-09
qwen2.5-1.5b 28.7 46.8 6.7 1.5B 2026-09-08
precog-123b-v1a-q4 6.7 – 6.7 β€” 2026-09-09
  • Scores on this suite are percentages (0 to 100), not out of 20 like the current board. Total combines Chat and Code; # ranks by Total.
  • A model with a dash in Chat or Code only ran one of the two, so its Total rests on that part alone.

MakoBench is free and has no ads. Running it costs power, hardware and our time.

β˜• Donate: Help keep us going

Request a model Want a model tested? Search the list, add your +1, or send a new one.