Skip to main content

FrontierMath v2 Tiers 1-3 (FrontierMath v2 (Tiers 1-3))

Epoch AI's corrected v2 core FrontierMath suite of private advanced mathematics problems. Models can reason iteratively and use Python; scores are pass rates on the private set.

Data verified

Top models on FrontierMath v2 (Tiers 1-3) — July 9, 2026

As of July 9, 2026, GPT-5.6 Sol leads the FrontierMath v2 (Tiers 1-3) leaderboard with 89.000% , followed by GPT-5.6 Terra (84.900%) and GPT-5.6 Luna (78.600%).

53 modelsMath30% of category scoreCurrentUpdated July 9, 2026

According to BenchLM.ai, GPT-5.6 Sol leads the FrontierMath v2 (Tiers 1-3) benchmark with a score of 89.000%, followed by GPT-5.6 Terra (84.900%) and GPT-5.6 Luna (78.600%). There is significant spread across the leaderboard, making this benchmark effective at differentiating model capabilities.

53 models have been evaluated on FrontierMath v2 (Tiers 1-3). The benchmark falls in the Math category. This category carries a 5% weight in BenchLM.ai's overall scoring system. Within that category, FrontierMath v2 (Tiers 1-3) contributes 30% of the category score, so strong performance here directly affects a model's overall ranking.

About FrontierMath v2 (Tiers 1-3)

Year

2026

Tasks

295 private advanced mathematics problems

Format

Python-enabled iterative mathematical problem solving

Difficulty

From olympiad-plus to early research mathematics

After Epoch AI's June 2026 correction, the FrontierMath v2 private core contains 295 Tiers 1-3 problems. BenchLM selects the highest published reasoning-effort result for each model and keeps this core score distinct from Tier 4.

BenchLM freshness & provenance

Version

FrontierMath v2 (Tiers 1-3) 2026

Refresh cadence

Quarterly

Staleness state

Current

Question availability

Public benchmark set

Current

BenchLM uses freshness metadata to decide whether a benchmark should still be treated as a strong differentiator, a benchmark to watch, or a display-only reference. For the full scoring policy, see the BenchLM methodology page.

Leaderboard (53 models)

1
GPT-5.6 Solgpt-5-6-sol
89.000%
2
GPT-5.6 Terragpt-5-6-terra
84.900%
3
GPT-5.6 Lunagpt-5-6-luna
78.600%
4
GPT-5.5gpt-5-5
51.700%
5
GPT-5.5 Progpt-5-5-pro
51.000%
6
GPT-5.4 Progpt-5-4-pro
50.000%
7
GPT-5.4gpt-5-4
47.600%
8
Claude Opus 4.8claude-opus-4-8
47.241%
9
Claude Opus 4.7claude-opus-4-7
43.793%
10
Claude Opus 4.6claude-opus-4-6
40.700%
11
GPT-5.2gpt-5-2
40.700%
12
Muse Sparkmuse-spark
39.000%
13
Gemini 3.5 Flashgemini-3-5-flash
38.966%
14
Kimi K2.6kimi-2-6
38.966%
15
Gemini 3 Progemini-3-pro
37.600%
16
Gemini 3.1 Progemini-3-1-pro
36.900%
17
Gemini 3 Flashgemini-3-flash
35.640%
18
GLM-5.1glm-5-1
33.448%
19
Claude Sonnet 4.6claude-sonnet-4-6
32.400%
20
GPT-5.1gpt-5-1
31.034%
21
GPT-5.4 minigpt-5-4-mini
28.280%
22
Kimi K2.5kimi-k2-5
27.900%
23
GPT-5 minigpt-5-mini
27.241%
24
Qwen3.6 Plusqwen3-6-plus
26.207%
25
GPT-5.4 nanogpt-5-4-nano
25.860%
26
o4-mini (high)o4-mini-high
24.828%
27
Qwen 3.6 Max (preview)qwen3-6-max-preview
23.103%
28
DeepSeek V3.2deepseek-v3-2
22.100%
29
Kimi K2kimi-k2
21.404%
30
Qwen3.5 Plusqwen3-5-plus
21.034%
31
Claude Opus 4.5claude-opus-4-5
20.690%
32
Grok 4grok-4
19.655%
33
18.685%
34
GLM-5glm-5
16.434%
35
Gemini 2.5 Progemini-2-5-pro
14.138%
36
Claude Sonnet 4.5claude-sonnet-4-5
13.495%
37
9.310%
38
Qwen3 235B 2507 (Reasoning)qwen3-235b-2507-reasoning
8.481%
39
GPT-5 nanogpt-5-nano
8.276%
40
Qwen3.5 Flashqwen3-5-flash
6.207%
41
Claude Haiku 4.5claude-haiku-4-5
5.903%
42
GPT-4.1gpt-4-1
5.517%
43
Gemini 2.5 Flashgemini-2-5-flash
4.844%
44
GPT-4.1 minigpt-4-1-mini
4.483%
45
GLM-4.6glm-4-6
3.819%
46
Grok 3 [Beta]grok-3-beta
3.793%
47
GLM-4.7glm-4-7
2.439%
48
Claude 3.5 Sonnetclaude-3-5-sonnet
2.069%
49
DeepSeek V3deepseek-v3
1.724%
50
GPT-4.1 nanogpt-4-1-nano
1.034%
51
Llama 4 Maverickllama-4-maverick
0.690%
52
GPT-4ogpt-4o
0.345%
53
Llama 4 Scoutllama-4-scout
0.000%

FAQ

What does FrontierMath v2 (Tiers 1-3) measure?

Epoch AI's corrected v2 core FrontierMath suite of private advanced mathematics problems. Models can reason iteratively and use Python; scores are pass rates on the private set.

Which model scores highest on FrontierMath v2 (Tiers 1-3)?

GPT-5.6 Sol by OpenAI currently leads with a score of 89.000% on FrontierMath v2 (Tiers 1-3).

How many models are evaluated on FrontierMath v2 (Tiers 1-3)?

53 AI models have been evaluated on FrontierMath v2 (Tiers 1-3) on BenchLM.

Last updated: July 9, 2026 · BenchLM version FrontierMath v2 (Tiers 1-3) 2026

The AI models change fast. We track them for you.

For engineers, researchers, and the plain curious — a weekly brief on new models, ranking shifts, and pricing changes.

Free. No spam. Unsubscribe anytime.