Field study / Version 2 · Archive
Version 2 · Archive

ChronoBench

Measuring how far different language models can progress inChrono Trigger using an autonomous vision-based game agent with procedural skills, curated notes, and event-driven strategic reasoning.

Last updated: 2026-04-19 · 200-cycle budget · v2 harness · 6 runs frozen.

These runs used the v2 harness. For current runs on the evidence-validating v3 harness, seeVersion 3.

Leaderboard

ModelProviderLast Checkpoint123456CyclesStuckSkillsSuperegoTokens Inms/cycleEst. CostDate
google/gemini-3-flash-previewOpenRoutermarle_met24444865——200*597244,200,24013,907$2.292026-04-18
xiaomi/mimo-v2-omniOpenRouterfair_entered154278———193317351,786,89960,361—2026-04-19
qwen/qwen3.6-35b-a3bLocalfair_entered233365———200*8—341,992,17533,893Local · $02026-04-18
z-ai/glm-5v-turboOpenRouterfair_entered263698———200*16—281,932,25947,551$3.082026-04-18
google/gemma-4-26b-a4bOpenRouterhouse_exit1764————200*57—262,159,50025,968—2026-04-19
google/gemma-4-e4bOpenRouterhouse_exit1663————200*1512422,338,56322,846—2026-04-18

* = cycle budget exhausted

Cycles per checkpoint

022436586108CyclesLeft Bedroomgoogle/gemini-3-flash-preview: 2424xiaomi/mimo-v2-omni: 1515qwen/qwen3.6-35b-a3b: 2323z-ai/glm-5v-turbo: 2626google/gemma-4-26b-a4b: 1717google/gemma-4-e4b: 1616Exited Housegoogle/gemini-3-flash-preview: 4444xiaomi/mimo-v2-omni: 4242qwen/qwen3.6-35b-a3b: 3333z-ai/glm-5v-turbo: 3636google/gemma-4-26b-a4b: 6464google/gemma-4-e4b: 6363Reached the Fairgoogle/gemini-3-flash-preview: 4848xiaomi/mimo-v2-omni: 7878qwen/qwen3.6-35b-a3b: 6565z-ai/glm-5v-turbo: 9898Met Marlegoogle/gemini-3-flash-preview: 6565Reached TelepodTime Traveledgoogle/gemini-3-flash-previewxiaomi/mimo-v2-omniqwen/qwen3.6-35b-a3bz-ai/glm-5v-turbogoogle/gemma-4-26b-a4bgoogle/gemma-4-e4b

Estimated cost per checkpoint

$0.00$0.40$0.80$1.20$1.60$2.00Cost ($)Left Bedroomgoogle/gemini-3-flash-preview: $0.27$0.27qwen/qwen3.6-35b-a3b: $0.00z-ai/glm-5v-turbo: $0.40$0.40Exited Housegoogle/gemini-3-flash-preview: $0.50$0.50qwen/qwen3.6-35b-a3b: $0.00z-ai/glm-5v-turbo: $0.56$0.56Reached the Fairgoogle/gemini-3-flash-preview: $0.55$0.55qwen/qwen3.6-35b-a3b: $0.00z-ai/glm-5v-turbo: $1.51$1.51Met Marlegoogle/gemini-3-flash-preview: $0.74$0.74Reached TelepodTime Traveledgoogle/gemini-3-flash-previewxiaomi/mimo-v2-omniqwen/qwen3.6-35b-a3bz-ai/glm-5v-turbogoogle/gemma-4-26b-a4bgoogle/gemma-4-e4b