Which model leads livecodebench_hard?

livecodebench_hard Leaderboard — Model Scores

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

The field · 17 measured

1.88666

17 models measured, most on the third-party eval harness. O3 tops the board at 66.

17 measured·Lifecycle Updated Oct 8, 2026Score basis

#◆Vendor$/M in$/M out

Leader — ignites whiteField — fill cools with scoreIndependently verified only17 of 17Bars show 1.886–66 · ◆ solid = first-party or better · outlined = arm's-length

01O3OpenAI66VENDOR$2.0/M$8.0/M$0.076

02Gemini 2.5 Pro+1 altGoogle59.429$1.3/M$10.0/M$0.095

03OpenReasoning-Nemotron-32BNVIDIA57.514———

04EXAONE 4.0 32BLG AI56.286VENDOR———

05Qwen3 235B A22BAlibaba54$0.70/M$2.8/M$0.032

06OpenCodeReasoning-Nemotron-1.1-32BNVIDIA52.771VENDOR———

07Grok 3 Mini ReasoningSpaceXAI51.143VENDOR$0.30/M$0.50/M$0.008

08O4 Mini+2 altsOpenAI48.857VENDOR$1.1/M$4.4/M$0.056

09Gemini 2.5 Flash+1 altGoogle48.286$0.30/M$2.5/M$0.029

10O3 Mini+2 altsOpenAI47.429$1.1/M$4.4/M$0.058

11Claude Opus 4Anthropic33.143$15.0/M$75.0/M$1.358

12Claude Sonnet 4Anthropic32.571VENDOR$3.0/M$15.0/M$0.276

13DeepSeek-V3DeepSeek14.2$0.24/M$0.90/M$0.040

14GPT-4oOpenAI4.457VENDOR$5.0/M$15.0/M$2.244

15GPT-4o miniOpenAI4.343$0.15/M$0.60/M$0.086

16GPT-4 TurboOpenAI4$10.0/M$30.0/M$5.000

17Claude 3 HaikuAnthropic1.886VENDOR$0.25/M$1.3/M$0.398

◆ 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: 17 models scored · 17 independently verified · 0 with source disagreement. How these tiers are assigned