Version 3 · Current

One model reached the End of Time.

google/gemini-3-flash-preview is the only agent — of 14 runs across 13 models — to finish the opening and time-travel to 600 AD.

8 / 8 checkpoints Time-traveled · cycle 182 $2.44 283 skill injections 12.2s / cycle
Replay the winning run Updated 2026-06-14 · 200-cycle budget

ChronoBench measures how far language models progress in Chrono Trigger using a vision-based agent with evidence-gated checkpoints and a first-class exploration track.

How it works Methodology View v2 archive View v1 archive

Top models

The podium

1st OpenRouter
Final state — google/gemini-3-flash-preview

google/gemini-3-flash-preview

Time Traveled

8/8 checkpoints $2.44

Finished the game

2nd OpenRouter
Final state — qwen/qwen3.7-plus

qwen/qwen3.7-plus

Reached Telepod

7/8 checkpoints $1.05
3rd Local
Final state — google/gemma-4-26b-a4b

google/gemma-4-26b-a4b

Met Marle

4/8 checkpoints Free

Full standings

Leaderboard

Best run per model, ranked by furthest checkpoint reached (ties broken by fewer cycles).

1

google/gemini-3-flash-preview

OpenRouter

8/8 · Time Traveled · finished

Cycles

200*

Stuck

17%

Tok/cp

569k

Cost

$2.44

Sec

2/10

2

qwen/qwen3.7-plus

OpenRouter

7/8 · Reached Telepod

Cycles

200*

Stuck

29%

Tok/cp

373k

Cost

$1.05

Sec

0/10

3

google/gemma-4-26b-a4b

Local

4/8 · Met Marle

Cycles

200*

Stuck

20%

Tok/cp

697k

Cost

Free

Sec

1/10

4

google/gemma-4-e4b

Local

3/8 · Reached the Fair

Cycles

200*

Stuck

59%

Tok/cp

977k

Cost

Free

Sec

0/10

5

x-ai/grok-4.3

OpenRouter

3/8 · Reached the Fair

Cycles

200*

Stuck

15%

Tok/cp

808k

Cost

$3.64

Sec

1/10

6

openai/gpt-5.4-nano

Local

2/8 · Exited House

Cycles

200*

Stuck

51%

Tok/cp

1.25M

Cost

Free

Sec

0/10

7

qwen/qwen3.6-35b-a3b

OpenRouter

2/8 · Exited House

Cycles

200*

Stuck

7%

Tok/cp

1.27M

Cost

$0.00

Sec

0/10

8

google/gemini-3.1-flash-lite

OpenRouter

2/8 · Exited House

Cycles

200*

Stuck

13%

Tok/cp

2.33M

Cost

$1.24

Sec

1/10

9

mistralai/mistral-medium-3-5

OpenRouter

2/8 · Exited House

Cycles

200*

Stuck

9%

Tok/cp

1.30M

Cost

$4.27

Sec

1/10

10

stepfun/step-3.7-flash

OpenRouter

2/8 · Exited House

Cycles

200*

Stuck

13%

Tok/cp

1.33M

Cost

$1.00

Sec

0/10

11

moonshotai/kimi-k2.7-code

OpenRouter

1/8 · Left Bedroom

Cycles

136

Stuck

7%

Tok/cp

1.71M

Cost

$1.96

Sec

1/10

12

moonshotai/kimi-k2.6

OpenRouter

0/8 · No checkpoints

Cycles

12

Stuck

0%

Tok/cp

—

Cost

$0.18

Sec

0/10

13

minimax/minimax-m3

OpenRouter

0/8 · No checkpoints

Cycles

200*

Stuck

29%

Tok/cp

—

Cost

$1.16

Sec

0/10

Progress shape

Cycles per checkpoint

0 40 80 121 161 201 Cycles Left Bedroom google/gemini-3-flash-preview: 14 qwen/qwen3.7-plus: 14 google/gemma-4-26b-a4b: 23 google/gemma-4-e4b: 19 x-ai/grok-4.3: 18 openai/gpt-5.4-nano: 48 qwen/qwen3.6-35b-a3b: 3 google/gemini-3.1-flash-lite: 18 mistralai/mistral-medium-3-5: 18 stepfun/step-3.7-flash: 15 moonshotai/kimi-k2.7-code: 24 Exited House google/gemini-3-flash-preview: 45 qwen/qwen3.7-plus: 35 google/gemma-4-26b-a4b: 40 google/gemma-4-e4b: 34 x-ai/grok-4.3: 18 openai/gpt-5.4-nano: 58 qwen/qwen3.6-35b-a3b: 5 google/gemini-3.1-flash-lite: 33 mistralai/mistral-medium-3-5: 18 stepfun/step-3.7-flash: 41 Reached the Fair google/gemini-3-flash-preview: 50 50 qwen/qwen3.7-plus: 83 83 google/gemma-4-26b-a4b: 52 52 google/gemma-4-e4b: 119 119 x-ai/grok-4.3: 117 117 openai/gpt-5.4-nano: 134 134 Met Marle google/gemini-3-flash-preview: 76 76 qwen/qwen3.7-plus: 131 131 google/gemma-4-26b-a4b: 136 136 Telepod Demo Announced google/gemini-3-flash-preview: 94 94 qwen/qwen3.7-plus: 151 151 Passed Candy Gate google/gemini-3-flash-preview: 104 104 qwen/qwen3.7-plus: 165 165 Reached Telepod google/gemini-3-flash-preview: 107 107 qwen/qwen3.7-plus: 180 180 Time Traveled google/gemini-3-flash-preview: 182 182 google/gemini-3-flash-preview qwen/qwen3.7-plus google/gemma-4-26b-a4b google/gemma-4-e4b x-ai/grok-4.3 openai/gpt-5.4-nano qwen/qwen3.6-35b-a3b google/gemini-3.1-flash-lite mistralai/mistral-medium-3-5 stepfun/step-3.7-flash moonshotai/kimi-k2.7-code

Cost frontier

Estimated cost per checkpoint

$0.00 $0.60 $1.20 $1.80 $2.40 $3.00 Cost ($) Left Bedroom google/gemini-3-flash-preview: $0.17 qwen/qwen3.7-plus: $0.07 x-ai/grok-4.3: $0.33 google/gemini-3.1-flash-lite: $0.11 mistralai/mistral-medium-3-5: $0.38 stepfun/step-3.7-flash: $0.08 moonshotai/kimi-k2.7-code: $0.35 Exited House google/gemini-3-flash-preview: $0.55 $0.55 qwen/qwen3.7-plus: $0.18 x-ai/grok-4.3: $0.33 google/gemini-3.1-flash-lite: $0.20 mistralai/mistral-medium-3-5: $0.38 $0.38 stepfun/step-3.7-flash: $0.21 Reached the Fair google/gemini-3-flash-preview: $0.61 $0.61 qwen/qwen3.7-plus: $0.44 $0.44 x-ai/grok-4.3: $2.13 $2.13 Met Marle google/gemini-3-flash-preview: $0.93 $0.93 qwen/qwen3.7-plus: $0.69 $0.69 Telepod Demo Announced google/gemini-3-flash-preview: $1.15 $1.15 qwen/qwen3.7-plus: $0.79 $0.79 Passed Candy Gate google/gemini-3-flash-preview: $1.27 $1.27 qwen/qwen3.7-plus: $0.87 $0.87 Reached Telepod google/gemini-3-flash-preview: $1.31 $1.31 qwen/qwen3.7-plus: $0.95 $0.95 Time Traveled google/gemini-3-flash-preview: $2.22 $2.22 google/gemini-3-flash-preview qwen/qwen3.7-plus google/gemma-4-26b-a4b google/gemma-4-e4b x-ai/grok-4.3 openai/gpt-5.4-nano qwen/qwen3.6-35b-a3b google/gemini-3.1-flash-lite mistralai/mistral-medium-3-5 stepfun/step-3.7-flash moonshotai/kimi-k2.7-code

Beyond the critical path

Exploration track

Secondary checkpoints reward engaging with the world instead of rushing the plot — they don't affect primary rank. Almost every model speed-runs the story and ignores them, so coverage here is a different kind of intelligence signal. Cells marked rare were found by only one model.

Model allowance gato soda race_bet melchior cat_returned lunch_eaten pendant_sold bekkler_lab wait_battle_mode Total
google/gemini-3-flash-preview
25
— — — — — — — —
1
2/10
qwen/qwen3.7-plus — — — — — — — — — — 0/10
google/gemma-4-26b-a4b — — — —
81
— — — — — 1/10
google/gemma-4-e4b — — — — — — — — — — 0/10
x-ai/grok-4.3
54
— — — — — — — — — 1/10
openai/gpt-5.4-nano — — — — — — — — — — 0/10
qwen/qwen3.6-35b-a3b — — — — — — — — — — 0/10
google/gemini-3.1-flash-lite
26
— — — — — — — — — 1/10
mistralai/mistral-medium-3-5
137
— — — — — — — — — 1/10
stepfun/step-3.7-flash — — — — — — — — — — 0/10
moonshotai/kimi-k2.7-code
77
— — — — — — — — — 1/10
moonshotai/kimi-k2.6 — — — — — — — — — — 0/10
minimax/minimax-m3 — — — — — — — — — — 0/10

Cells show the first cycle each secondary checkpoint was confirmed. Hover for the full label.

Every session

All runs

Model Last primary Cycles Secondary Stuck Date
google/gemini-3-flash-preview 1000ad_left 200* 2/10 33 2026-04-20
google/gemini-3-flash-preview telepod_reached 200* 1/10 27 2026-04-19
qwen/qwen3.7-plus telepod_reached 200* — 58 2026-06-13
google/gemma-4-26b-a4b marle_met 200* 1/10 39 2026-04-20
google/gemma-4-e4b fair_entered 200* — 118 2026-04-20
x-ai/grok-4.3 fair_entered 200* 1/10 30 2026-05-09
openai/gpt-5.4-nano house_exit 200* — 102 2026-04-19
qwen/qwen3.6-35b-a3b house_exit 200* — 13 2026-04-20
google/gemini-3.1-flash-lite house_exit 200* 1/10 25 2026-05-08
mistralai/mistral-medium-3-5 house_exit 200* 1/10 18 2026-05-09
stepfun/step-3.7-flash house_exit 200* — 26 2026-06-14
moonshotai/kimi-k2.7-code bedroom_exit 136 1/10 10 2026-06-13
moonshotai/kimi-k2.6 — 12 — 0 2026-04-20
minimax/minimax-m3 — 200* — 58 2026-06-13