Which model leads frontiermath_tier_4_v1?

Q: Which model leads frontiermath_tier_4_v1?

GPT-5.5 Pro tops the board at 39.6

frontiermath_tier_4_v1 Leaderboard — Model Scores

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

The field · 52 measured

039.6

52 models measured, most on the third-party eval harness. GPT-5.5 Pro tops the board at 39.6.

52 measured·Lifecycle Updated Oct 29, 2026Score basis

#◆Vendor$/M in$/M out

Leader — ignites whiteField — fill cools with scoreIndependently verified only52 of 52Bars show 0–39.6 · ◆ solid = first-party or better · outlined = arm's-length

01GPT-5.5 ProOpenAI39.6$30.0/M$180.0/M$2.652

02GPT-5.4 ProOpenAI37.5$30.0/M$180.0/M$2.800

03GPT-5.5OpenAI35.4$5.0/M$30.0/M$0.494

04GPT-5.2 ProOpenAI31.3$21.0/M$168.0/M$3.019

05Claude Opus 4.8Anthropic31.25$5.0/M$25.0/M$0.480

06GPT-5.4OpenAI27.1$2.5/M$15.0/M$0.323

07Claude Opus 4.7Anthropic22.92$5.0/M$25.0/M$0.654

08Claude Opus 4.6+2 altsAnthropic22.9$5.0/M$25.0/M$0.655

09GPT-5.2+4 altsOpenAI18.8$1.8/M$14.0/M$0.419

10Gemini 3 ProGoogle18.75$2.0/M$12.0/M$0.373

11Gemini 3.1 ProGoogle16.7$2.0/M$12.0/M$0.419

12GPT-5 ProOpenAI14.6$15.0/M$120.0/M$4.623

13Muse SparkMeta14.6———

14Kimi K2.6Moonshot14.58$0.95/M$4.0/M$0.170

15Gemini 3.5 FlashGoogle14.58$1.5/M$9.0/M$0.360

16GLM-5.1Z.ai12.5$1.4/M$4.4/M$0.231

17GPT-5+1 altOpenAI12.5$1.3/M$10.0/M$0.450

18GPT-5.1+2 altsOpenAI12.5$1.3/M$10.0/M$0.450

19Gemini 2.5 Deep ThinkGoogle10.4———

20Qwen3.6 PlusAlibaba8.33$0.50/M$3.0/M$0.210

21Claude Sonnet 4.6Anthropic8.3$3.0/M$15.0/M$1.084

22GPT-5.4 nanoOpenAI6.25$0.20/M$1.3/M$0.116

23O4 Mini+2 altsOpenAI6.25$1.1/M$4.4/M$0.440

24GPT-5 mini+2 altsOpenAI6.25$0.25/M$2.0/M$0.180

25Kimi K2.5Moonshot4.2$0.50/M$2.5/M$0.357

26Gemini 3 FlashGoogle4.17$0.50/M$3.0/M$0.420

27Gemini 2.5 FlashGoogle4.17$0.30/M$2.5/M$0.336

28Claude Sonnet 4.5+1 altAnthropic4.17$3.0/M$15.0/M$2.158

29O3 MiniOpenAI4.17$1.1/M$4.4/M$0.659

30Qwen3.6 Max PreviewAlibaba4.17———

31Gemini 2.5 ProGoogle4.17$1.3/M$10.0/M$1.349

32Claude Opus 4.1Anthropic4.17$15.0/M$75.0/M$10.791

33Claude Opus 4.5+1 altAnthropic4.17$5.0/M$25.0/M$3.597

34Claude Opus 4Anthropic4.17$15.0/M$75.0/M$10.791

35GLM-4.6Z.ai2.13$0.57/M$2.2/M$0.651

36DeepSeek-V3.2DeepSeek2.1$0.26/M$0.38/M$0.152

37GLM-5Z.ai2.1$1.0/M$3.2/M$1.000

38GPT-5.4 miniOpenAI2.08$0.75/M$4.5/M$1.262

39Qwen3.5 PlusAlibaba2.08$0.40/M$2.4/M$0.673

40Grok 4SpaceXAI2.08$3.0/M$15.0/M$4.327

41Claude Haiku 4.5Anthropic2.08$1.0/M$5.0/M$1.442

42Grok 4 HeavySpaceXAI2.08———

43O3OpenAI2.08$2.0/M$8.0/M$2.404

44GPT-5 nano+1 altOpenAI2.08$0.05/M$0.40/M$0.108

45GPT-4.1OpenAI0$2.0/M$8.0/M—

46Claude Sonnet 4Anthropic0$3.0/M$15.0/M—

47GLM-4.7Z.ai0$0.53/M$2.1/M—

48Qwen 3.6 FlashAlibaba0$0.25/M$1.5/M—

49Qwen3.5 FlashAlibaba0$0.10/M$0.40/M—

50Claude 3.5 Sonnet+1 altAnthropic0$3.0/M$15.0/M—

51Kimi K2 ThinkingMoonshot0$0.60/M$2.5/M—

52Grok 3SpaceXAI0$3.0/M$15.0/M—

◆ 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: 52 models scored · 52 independently verified · 0 with source disagreement. How these tiers are assigned