Which model leads SWT-bench?

SWT-bench Leaderboard — Model Scores

Pro Mode — raw scores, one benchmark under the lensthird-party eval harness

The field · 7 measured

32.880.7

7 models measured, most on the third-party eval harness. GPT-5.2 tops the board at 80.7.

7 measured·Lifecycle Updated Oct 8, 2026Score basis

#◆Vendor$/M in$/M out

Leader — ignites whiteField — fill cools with scoreIndependently verified only0 of 7Bars show 32.8–80.7 · ◆ solid = first-party or better · outlined = arm's-length

01GPT-5.2OpenAI80.7$1.8/M$14.0/M$0.098

02Claude Opus 4.5Anthropic80.2$5.0/M$25.0/M$0.187

03Gemini 3 ProGoogle79.7$2.0/M$12.0/M$0.088

04Claude Sonnet 4.5Anthropic69.5$3.0/M$15.0/M$0.129

05MiniMax M2.1+1 altMiniMax69.3VENDOR$0.30/M$1.2/M$0.011

06DeepSeek-V3.2DeepSeek62$0.28/M$0.42/M$0.006

07MiniMax M2MiniMax32.8$0.30/M$1.2/M$0.023

◆ 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: 7 models scored · 0 independently verified · 5 vendor cross-reference · 2 vendor-reported · 0 with source disagreement. How these tiers are assigned