Benchmark · Reasoning and knowledge
FrontierMath Tier 4
The hardest FrontierMath problems, an expansion set of 43 in version 2, each taking a specialist researcher days. Same method as FrontierMath's Tiers 1-3, run by Epoch AI on the private set.
- Score
- % of problems solved, higher is better
- Run by
- Epoch AI
- Status
- Current, newest result 19 November 2026
- Updated
- as models are released
- Ceiling
- The best result is within five points of a perfect score, so the test no longer tells the top models apart
- Data
- the published file, CC BY 4.0, read 20 November 2026
Results: Tier 4 v2 (2026-06-12)
32 models we list, each at its best setting, best first.
| # | Model | Score | Range | Who ran it | Setting | Date | |
|---|---|---|---|---|---|---|---|
| 1 | Claude Fable 5.2 A | 100.0% | Independent | effort max | 2026-11-06 | source | |
| 1 | GPT-6.2 Astra AA | 100.0% | Independent | effort max | 2026-10-27 | source | |
| 1 | GPT-6.1 Sol AA | 100.0% | Independent | effort max | 2026-09-29 | source | |
| 4 | GPT-6 Astra A | 97.6% | Independent | effort high | 2026-08-30 | source | |
| 5 | Gemini 4 Neon B | 97.6% | Independent | effort xhigh | 2026-11-19 | source | |
| 6 | Gemini 4 Carbon B | 95.1% | Independent | effort high | 2026-11-02 | source | |
| 7 | Claude Opus 5.5 A | 95.0% | Independent | effort max | 2026-09-22 | source | |
| 8 | GPT-6.1 Terra AA | 92.7% | Independent | effort max | 2026-10-26 | source | |
| 9 | GPT-6 Sol A | 90.0% | Independent | effort max | 2026-09-22 | source | |
| 10 | Gemini 4 Hydrogen B | 87.8% | Independent | effort xhigh | 2026-11-13 | source | |
| 10 | Claude Fable 5.1 A | 87.8% | Independent | effort max | 2026-09-01 | source | |
| 12 | Claude Sonnet 5.5 A | 80.5% | Independent | effort max | 2026-09-29 | source | |
| 13 | Gemini 4 Argon B | 78.0% | Independent | effort high | 2026-10-13 | source | |
| 14 | Muse Wonder D | 70.7% | Independent | effort max | 2026-10-12 | source | |
| 15 | Kimi K3.1 Pro D | 65.8% | Independent | effort max | 2026-11-06 | source | |
| 16 | GPT-6 Luna A | 56.1% | Independent | effort max | 2026-09-22 | source | |
| 17 | DeepSeek V4.1 Flash C | 46.3% | Independent | effort max | 2026-10-19 | source | |
| 17 | Claude Haiku 5.5 A | 46.3% | Independent | effort max | 2026-10-07 | source | |
| 17 | Muse Spark 1.3 D | 46.3% | Independent | effort max | 2026-09-18 | source | |
| 20 | Kimi K3 B | 39.0% | Independent | effort max | 2026-07-17 | source | |
| 21 | Muse Spark Open C | 31.7% | Independent | effort xhigh | 2026-10-15 | source | |
| 22 | GLM-5.3 B | 29.3% | Independent | effort max | 2026-08-25 | source | |
| 23 | Gemini 3.1 Pro (preview) AA | 26.8% | Independent | 2026-06-11 | source | ||
| 24 | DeepSeek-V4-Flash-0731 C | 24.4% | Independent | effort max | 2026-08-02 | source | |
| 25 | Gemini 3.8 Flash AA | 21.9% | Independent | effort high | 2026-09-02 | source | |
| 26 | Grok 4.7 BB | 17.1% | Independent | effort xhigh | 2026-09-22 | source | |
| 26 | GLM-5.3-Flash B | 17.1% | Independent | effort max | 2026-08-27 | source | |
| 26 | Inkling-Small B | 17.1% | Independent | effort xhigh | 2026-08-14 | source | |
| 29 | Grok 4.3 BB | 14.6% | Independent | effort high | 2026-06-17 | source | |
| 30 | Inkling B | 4.9% | Independent | effort xhigh | 2026-08-06 | source | |
| 31 | DeepSeek V4 Pro D | 2.4% | Independent | effort max | 2026-06-17 | source | |
| 32 | Gemini 3.5 Flash-Lite AA | 0.0% | Independent | effort high | 2026-08-02 | source |
Every setting, 39 rows
| Model | As the leaderboard names it | Score | Who ran it | Setting | Date | |
|---|---|---|---|---|---|---|
| Claude Fable 5.2 | claude-fable-5-2 | 100.0% | Independent | effort max | 2026-11-06 | source |
| GPT-6.2 Astra | gpt-6.2-astra | 100.0% | Independent | effort max | 2026-10-27 | source |
| GPT-6.1 Sol | gpt-6.1-sol | 100.0% | Independent | effort max | 2026-09-29 | source |
| GPT-6 Astra | gpt-6-astra | 97.6% | Independent | effort high | 2026-08-30 | source |
| GPT-6 Astra | gpt-6-astra | 97.6% | Independent | effort max | 2026-08-30 | source |
| GPT-6 Astra | gpt-6-astra | 97.6% | Independent | effort xhigh | 2026-08-30 | source |
| Gemini 4 Neon | gemini-4-neon | 97.6% | Independent | effort xhigh | 2026-11-19 | source |
| GPT-6 Astra | gpt-6-astra | 97.6% | Independent | effort medium | 2026-08-30 | source |
| Gemini 4 Carbon | gemini-4-carbon | 95.1% | Independent | effort high | 2026-11-02 | source |
| Claude Opus 5.5 | claude-opus-5-5 | 95.0% | Independent | effort max | 2026-09-22 | source |
| GPT-6.1 Terra | gpt-6.1-terra | 92.7% | Independent | effort max | 2026-10-26 | source |
| GPT-6 Sol | gpt-6-sol | 90.0% | Independent | effort max | 2026-09-22 | source |
| Gemini 4 Hydrogen | gemini-4-hydrogen | 87.8% | Independent | effort xhigh | 2026-11-13 | source |
| Claude Fable 5.1 | claude-fable-5-1 | 87.8% | Independent | effort max | 2026-09-01 | source |
| GPT-6 Astra | gpt-6-astra | 87.8% | Independent | effort low | 2026-08-30 | source |
| GPT-6 Astra | gpt-6-astra | 82.9% | Independent | effort none | 2026-08-30 | source |
| Claude Sonnet 5.5 | claude-sonnet-5-5 | 80.5% | Independent | effort max | 2026-09-29 | source |
| Gemini 4 Argon | gemini-4-argon | 78.0% | Independent | effort high | 2026-10-13 | source |
| Muse Wonder | muse-wonder-1.0 | 70.7% | Independent | effort max | 2026-10-12 | source |
| Kimi K3.1 Pro | kimi-k3.1-pro-together | 65.8% | Independent | effort max | 2026-11-06 | source |
| GPT-6 Luna | gpt-6-luna | 56.1% | Independent | effort max | 2026-09-22 | source |
| DeepSeek V4.1 Flash | deepseek-v4.1-flash-1017 | 46.3% | Independent | effort max | 2026-10-19 | source |
| Claude Haiku 5.5 | claude-haiku-5-5 | 46.3% | Independent | effort max | 2026-10-07 | source |
| Muse Spark 1.3 | muse-spark-1.3 | 46.3% | Independent | effort max | 2026-09-18 | source |
| Muse Spark 1.3 | muse-spark-1.3 | 41.5% | Independent | effort xhigh | 2026-09-16 | source |
| Kimi K3 | kimi-k3 | 39.0% | Independent | effort max | 2026-07-17 | source |
| Muse Spark Open | Muse-Spark-Open-816B-A42B | 31.7% | Independent | effort xhigh | 2026-10-15 | source |
| GLM-5.3 | glm-5.3 | 29.3% | Independent | effort max | 2026-08-25 | source |
| DeepSeek V4.1 Flash | deepseek-v4.1-flash | 26.8% | Independent | effort max | 2026-10-08 | source |
| Gemini 3.1 Pro (preview) | gemini-3.1-pro-preview | 26.8% | Independent | 2026-06-11 | source | |
| DeepSeek-V4-Flash-0731 | deepseek-v4-flash-0731 | 24.4% | Independent | effort max | 2026-08-02 | source |
| Gemini 3.8 Flash | gemini-3.8-flash | 21.9% | Independent | effort high | 2026-09-02 | source |
| Grok 4.7 | grok-4.7 | 17.1% | Independent | effort xhigh | 2026-09-22 | source |
| GLM-5.3-Flash | glm-5.3-flash | 17.1% | Independent | effort max | 2026-08-27 | source |
| Inkling-Small | inkling-small | 17.1% | Independent | effort xhigh | 2026-08-14 | source |
| Grok 4.3 | grok-4.3 | 14.6% | Independent | effort high | 2026-06-17 | source |
| Inkling | Inkling | 4.9% | Independent | effort xhigh | 2026-08-06 | source |
| DeepSeek V4 Pro | deepseek-v4-pro | 2.4% | Independent | effort max | 2026-06-17 | source |
| Gemini 3.5 Flash-Lite | gemini-3.5-flash-lite | 0.0% | Independent | effort high | 2026-08-02 | source |
Results: Tier 4 (2025-07-01)
1 model we list, each at its best setting, best first.
| # | Model | Score | Range | Who ran it | Setting | Date | |
|---|---|---|---|---|---|---|---|
| 1 | Gemini 3.1 Pro (preview) AA | 16.7% | Independent | 2026-02-19 | source |
A rank is among the models we list with a result, not a place on the whole leaderboard, and a model offered by several hosts is shown once, on its own listing. These results don't move our grades. Data: Epoch AI, Capabilities and benchmarking (CC BY 4.0), CC BY 4.0.
For companies
Do agents find, use and choose your tools?
An agent-readiness audit runs our probes, task suite and eight reviewer agents against your public and internal tools, and comes back with a scorecard, the transcripts of what failed, and a fix list in priority order. From $2,500, re-run included. We never take payment to move a rank. We do help companies earn one.