Vending-Bench 2
Can the agent run a business over a long horizon?
by Andon LabsAgents manage a simulated vending-machine business over a year-long horizon: ordering stock, setting prices, paying fees. Score is final net worth. Results by Andon Labs, displayed with attribution.
Net worth
↑ higher is better
Claude Opus 5
Anthropic
Net worth: $11,182 ± $2,094
Claude Opus 5$11,182 ± $2,094Anthropic
Net worth: $11,182 ± $2,094
Claude Opus 4.7
Anthropic
Net worth: $10,937 ± $1,181
Claude Opus 4.7$10,937 ± $1,181Anthropic
Net worth: $10,937 ± $1,181
GPT-5.6 Sol
OpenAI
Net worth: $9,619 ± $1,338
GPT-5.6 Sol$9,619 ± $1,338OpenAI
Net worth: $9,619 ± $1,338
GLM-5.2
Z.ai
Net worth: $8,314 ± $1,084
GLM-5.2$8,314 ± $1,084Z.ai
Net worth: $8,314 ± $1,084
Claude Opus 4.6
Anthropic
Net worth: $8,018 ± $1,367
Claude Opus 4.6$8,018 ± $1,367Anthropic
Net worth: $8,018 ± $1,367
GPT-5.5
OpenAI
Net worth: $7,524 ± $1,346
GPT-5.5$7,524 ± $1,346OpenAI
Net worth: $7,524 ± $1,346
GPT-5.6 Terra
OpenAI
Net worth: $7,343 ± $373
GPT-5.6 Terra$7,343 ± $373OpenAI
Net worth: $7,343 ± $373
Claude Sonnet 4.6
Anthropic
Net worth: $7,204 ± $722
Claude Sonnet 4.6$7,204 ± $722Anthropic
Net worth: $7,204 ± $722
Muse Spark 1.1
Meta
Net worth: $6,520 ± $1,016
Muse Spark 1.1$6,520 ± $1,016Meta
Net worth: $6,520 ± $1,016
Claude Sonnet 5
Anthropic
Net worth: $6,378 ± $784
Claude Sonnet 5$6,378 ± $784Anthropic
Net worth: $6,378 ± $784
results as of 2026-08-04 · full results at source ↗