Model Cards

Every ability card the arena has measured, one row each. Open a model to see its card and every board measurement behind it. A card is absolute, not relative to the field, which is why averaging cards is sound where merging leaderboards is not.

derived from the arena database on 2026-09-04

Board contestants

21 cards · each averaged across the boards it contested

One row per consolidated card, ordered by OVR. Deliberately unnumbered: these means span different sets of boards, so a place in this list would be the cross-workflow ranking the leaderboard refuses to publish.

ModelOVRGRDADHSYNPRCEFFCONArchetypeRange
gemini-3-6-flash93999899947986Sniper84–993 of 5 boards
gpt-5-6-terra90949089908592Sniper83–984 of 5 boards
deepseek-v4-pro88969096887489Sniper84–964 of 5 boards
grok-4-588999099974994Sniper73–984 of 5 boards
deepseek-v4-flash86969096895787Sniper75–924 of 5 boards
qwen-3-7-max84958799906178Playmaker75–944 of 5 boards
gpt-5-6-luna82879290935680Anchor68–935 of 5 boards
mimo-2-5-pro82938987836774Sniper65–984 of 5 boards
deepseek-v4-flash-vision81949999602496Anchor81–811 of 5 boards
gemini-3-5-flash78848889943383Anchor61–954 of 5 boards
kimi-2-778908872826666Sniper68–984 of 5 boards
glm-5-3-flash78899999701886Anchor78–781 of 5 boards
hunyuan-376938781803278Sniper69–824 of 5 boards
claude-sonnet-575868199865747Playmaker62–814 of 5 boards
glm-5-275858874796071Anchor64–854 of 5 boards
step-3-7-flash74838676885466Anchor68–814 of 5 boards
doubao-seed-2-1-pro73868887883460Anchor66–804 of 5 boards
longcat-2-072848487872968Playmaker46–834 of 5 boards
mimo-2-570809074846532Anchor62–794 of 5 boards
gemini-3-7-flash6759999964892Anchor67–671 of 5 boards
minimax-m364788060724250Anchor42–924 of 5 boards

Provisional

23 cards · measured, never ranked · grouped by model

Runs measured outside a contested field. An ability card is an absolute measurement — passed/total per axis, EFF against each workflow's own par — so it stays meaningful with no opponent. A rank does not: with no field, a first place measures nothing. Nothing here carries one, and nothing here reaches the leaderboard.

ModelOVRGRDADHSYNPRCEFFCONArchetypeRange
deepseek-v4-proRun #10482938984936072Sniper61–994 workflows · 2 trials
gpt-5-6-lunaRun #129 max84919495904789Playmaker73–985 workflows · 2 trials
gpt-5-6-lunaRun #130 low78819183876767Anchor49–965 workflows · 2 trials
glm-5-3-flashRun #131 max798592958825Playmaker69–935 workflows · 1 trial
hunyuan-3Run #129 xhigh80889377827850Anchor54–915 workflows
hunyuan-3Run #130 low79898979856363Sniper61–915 workflows
mimo-2-5Run #129 high70778081826352Anchor54–795 workflows · 2 trials
mimo-2-5Run #130 low76828279817170Sniper49–985 workflows
gemini-3-7-flashRun #113919996999660Sniper87–985 workflows · 1 trial
gemini-3-7-flashRun #129 xhigh83959299923189Playmaker77–965 workflows · 2 trials
gemini-3-7-flashRun #130 low86849499917686Playmaker73–985 workflows · 2 trials
minimax-m3Run #129 max72798679825558Anchor52–935 workflows · 2 trials
minimax-m3Run #130 low77808887805679Anchor61–985 workflows · 2 trials
gemini-3-8-flashRun #138 low869192999447Playmaker71–986 workflows · 1 trial
gemini-3-8-flashRun #134 xhigh7991979995489Playmaker71–846 workflows
grok-4-6Run #10481969099982186Playmaker74–914 workflows · 2 trials
glm-5-3Run #115879392959158Playmaker69–995 workflows · 1 trial
glm-5-3Run #132 max879091959064Playmaker72–995 workflows · 1 trial
qwen-3-8-flashRun #131 max788885998820Playmaker63–905 workflows · 1 trial
qwen-3-8-27bRun #129 max84918699895887Playmaker75–945 workflows · 2 trials
qwen-3-8-27bRun #130 low81888479817680Sniper53–955 workflows · 2 trials
deepseek-v4-flash-dsRun #129 max81919197885469Playmaker67–895 workflows · 2 trials
deepseek-v4-flash-dsRun #130 low84918895906186Playmaker73–925 workflows · 2 trials