Which model leads HealthBench Hard length-adjusted?

Q: Which model leads HealthBench Hard length-adjusted?

GPT-6 Astra tops the board at 36.6

HealthBench Hard length-adjusted Leaderboard

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

The field · 12 measured

25.436.6

12 models measured, most on the official card harness. GPT-6 Astra tops the board at 36.6.

12 measured·Lifecycle Updated Oct 8, 2026Score basis

#◆Vendor$/M in$/M out

Leader — ignites whiteField — fill cools with scoreIndependently verified only0 of 12Bars show 25.4–36.6 · ◆ solid = first-party or better · outlined = arm's-length

01GPT-6 Astra+1 altOpenAI36.6$10.0/M$50.0/M$0.820

02GPT-6.1 SolOpenAI36.2$2.0/M$10.0/M$0.166

03GPT-5OpenAI34.7$1.3/M$10.0/M$0.162

04GPT-5.2OpenAI34.3$1.8/M$14.0/M$0.230

05GPT-5.6 Sol+2 altsOpenAI33.1$4.0/M$20.0/M$0.363

06GPT-5.6 TerraOpenAI32.7$2.0/M$12.0/M$0.214

07GPT-5.6 LunaOpenAI32$0.20/M$1.2/M$0.022

08GPT-5.5OpenAI31.5$5.0/M$30.0/M$0.556

09GPT-6 LunaOpenAI31.4$0.10/M$0.50/M$0.010

10GPT-6 SolOpenAI30.1$2.0/M$10.0/M$0.199

11GPT-5.4OpenAI29.1$2.5/M$15.0/M$0.301

12GPT-5.1OpenAI25.4$1.3/M$10.0/M$0.221

◆ 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: 12 models scored · 0 independently verified · 12 vendor-reported · 0 with source disagreement. How these tiers are assigned