Which model leads SWE-Marathon?

SWE-Marathon Leaderboard — 11 Models Ranked

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

The field · 11 measured

446

11 models measured, most on the official card harness. Grok 4.7 tops the board at 46.

11 measured·Lifecycle Updated Oct 8, 2026Score basis

#◆Vendor$/M in$/M out

Leader — ignites whiteField — fill cools with scoreIndependently verified only0 of 11Bars show 4–46 · ◆ solid = first-party or better · outlined = arm's-length

01Grok 4.7SpaceXAI46$2.0/M$6.0/M$0.087

02GLM-5.3+2 altsZ.ai42.5$1.4/M$4.4/M$0.068

03GPT-5.6 Sol+1 altOpenAI42.53P$4.0/M$20.0/M$0.282

04Kimi K3+3 altsMoonshot42$0.58/M$12.3/M$0.153

05Claude Fable 5Anthropic353P$10.0/M$50.0/M$0.857

06Hy4Tencent31.9$0.83/M$2.5/M$0.052

07Grok4.5SpaceXAI29$2.0/M$6.0/M$0.138

08Claude Opus 4.8+4 altsAnthropic26$5.0/M$25.0/M$0.577

09GLM-5.2+6 altsZ.ai13$1.4/M$4.4/M$0.223

10GPT-5.5+2 altsOpenAI12$5.0/M$30.0/M$1.458

11Gemini 3.1 Pro+1 altGoogle4$2.0/M$12.0/M$1.750

◆ 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: 11 models scored · 0 independently verified · 5 vendor cross-reference · 6 vendor-reported · 1 with source disagreement. How these tiers are assigned