Field study / Version 3

The explorer / Version 3

Follow the evidence.

Every route has a story. Compare two to four runs from the same harness and inspect the checkpoints behind the numbers.

Enable JavaScript for interactive filtering and comparisons. All run reports remain available below.

Best run per model. Ranked by primary progress, then fewer cycles.
Rank / ModelJourneyCyclesInference costExploration
01 / google/gemini-3-flash-previewopenrouter8/8200Budget reached$2.442Secondary checkpoints
02 / qwen/qwen3.7-plusopenrouter7/8200Budget reached$1.050Secondary checkpoints
03 / google/gemma-4-26b-a4bLocal inference4/8200Budget reachedLocal · $01Secondary checkpoints
04 / google/gemma-4-e4bLocal inference3/8200Budget reachedLocal · $00Secondary checkpoints
05 / x-ai/grok-4.3openrouter3/8200Budget reached$3.641Secondary checkpoints
06 / openai/gpt-5.4-nanoLocal inference2/8200Budget reachedLocal · $00Secondary checkpoints
07 / qwen/qwen3.6-35b-a3bopenrouter2/8200Budget reached$0.000Secondary checkpoints
08 / google/gemini-3.1-flash-liteopenrouter2/8200Budget reached$1.241Secondary checkpoints
09 / mistralai/mistral-medium-3-5openrouter2/8200Budget reached$4.271Secondary checkpoints
10 / stepfun/step-3.7-flashopenrouter2/8200Budget reached$1.000Secondary checkpoints
11 / moonshotai/kimi-k2.7-codeopenrouter1/8136$1.961Secondary checkpoints
12 / moonshotai/kimi-k2.6openrouter0/812$0.180Secondary checkpoints
13 / minimax/minimax-m3openrouter0/8200Budget reached$1.160Secondary checkpoints