Which model leads Vending-Bench 2?

Q: Which model leads Vending-Bench 2?

Claude Opus 4.6 tops the board at 801,759

Vending-Bench 2 Leaderboard — Model Scores

Pro Mode — raw scores, one benchmark under the lensofficial card harness

The field · 9 measured

911.21801,759

9 models measured, most on the official card harness. Claude Opus 4.6 tops the board at 801,759.

9 measured·5 new this week·Lifecycle Updated Oct 9, 2026Score basis

#◆Vendor$/M in$/M out

Leader — ignites whiteField — fill cools with scoreIndependently verified only0 of 9Bars show 911.21–801,759 · ◆ solid = first-party or better · outlined = arm's-length

01Claude Opus 4.6+1 altAnthropic801,759$5.0/M$25.0/M$0.000

02GLM-5.1Z.ai563,441$1.4/M$4.4/M$0.000

03Gemini 3 ProGoogle547,816$2.0/M$12.0/M$0.000

04Gemini 3 Flash PreviewGoogle363,500$0.50/M$3.0/M$0.000

05GPT-5.4NEWOpenAI6,144.18$2.5/M$15.0/M$0.001

06GLM-5NEWZ.ai4,432.12$1.0/M$3.2/M$0.000

07Kimi K2.5NEWMoonshot1,198.46$0.45/M$2.3/M$0.001

08DeepSeek-V3.2NEWDeepSeek1,034$0.28/M$0.42/M$0.000

09Gemini 3.1 ProNEWGoogle911.21$2.0/M$12.0/M$0.008

◆ solid = independently verified · aggregator attested · vendor attributed — outlined = cross-reference · source-attributed. Dimmed ◆ = confidence inferred. “+N alts” = alternate disclosures in the drawer with why-demoted. Prices per million tokens; $/point = input price ÷ score. Every row opens its model dossier.

Verification: 9 models scored · 0 independently verified · 4 vendor cross-reference · 5 vendor-reported · 0 with source disagreement. How these tiers are assigned