Model Tracker
FrontierMath, tier 4 leaderboard
The models in the Model Tracker that Epoch AI has measured on FrontierMath, tier 4, ranked by the highest score of each: the name Epoch AI gives it, the setting it ran and the day.
En anglais
The models in the Model Tracker that Epoch AI has measured on FrontierMath, tier 4, ranked by the highest score of each: the name Epoch AI gives it, the setting it ran and the day. Epoch AI, ‘Capabilities & benchmarking’, epoch.ai/benchmarks, CC BY 4.0: Epoch AI’s own runs and its Capabilities Index.
- 1. GPT-6.1 Sol: 100.0% (± 0.0) — gpt-6.1-sol_max, max, 29 September 2026
- 1. Claude Fable 5.2: 100.0% (± 0.0) — claude-fable-5-2_max, max, 6 November 2026
- 3. GPT-6 Astra: 97.6% (± 2.4) — gpt-6-astra_high, high, 30 August 2026
- 4. Claude Opus 5.5: 95.0% (± 4.1) — claude-opus-5-5_max, max, 22 September 2026
- 5. Claude Fable 5: 90.2% (± 4.6) — claude-fable-5_max, max, 9 June 2026
- 6. GPT-6 Sol: 90.0% (± 5.2) — gpt-6-sol_max, max, 22 September 2026
- 7. Claude Fable 5.1: 87.8% (± 5.2) — claude-fable-5-1_max, max, 1 September 2026
- 8. GPT-5.6 Sol: 82.9% (± 5.9) — gpt-5.6-sol_max, max, 9 July 2026
- 9. Claude Sonnet 5.5: 80.5% (± 6.3) — claude-sonnet-5-5_max, max, 29 September 2026
- 10. Gemini 4 Argon: 78.0% (± 6.5) — gemini-4-argon_high, high, 13 October 2026
- 10. GPT-5.5 Pro: 78.0% (± 6.5) — gpt-5.5-pro_xhigh, xhigh, 12 June 2026
- 12. Claude Opus 5: 73.2% (± 7.0) — claude-opus-5_max, max, 24 July 2026
- 13. GPT-5.5: 72.5% (± 7.1) — gpt-5.5_xhigh, xhigh, 11 June 2026
- 14. Muse Wonder: 70.7% (± 7.2) — muse-wonder-1.0_max, max, 12 October 2026
- 14. GPT-5.6 Terra: 70.7% (± 7.2) — gpt-5.6-terra_max, max, 9 July 2026
- 16. GPT-5.6 Luna: 61.0% (± 7.7) — gpt-5.6-luna_max, max, 9 July 2026
- 17. GPT-5.4 Pro: 58.5% (± 7.8) — gpt-5.4-pro-2026-03-05_xhigh, xhigh, 13 June 2026
- 18. GPT-6 Luna: 56.1% (± 7.8) — gpt-6-luna_max, max, 22 September 2026
- 18. Claude Opus 4.8: 56.1% (± 7.8) — claude-opus-4-8_max, max, 10 June 2026
- 20. GPT-5.4: 49.0% (± 8.0) — gpt-5.4-2026-03-05_xhigh, xhigh, 11 June 2026
- 21. Claude Haiku 5.5: 46.3% (± 7.9) — claude-haiku-5-5_max, max, 7 October 2026
- 21. Qwen3.8-Max: 46.3% (± 7.9) — qwen3.8-max_xhigh, xhigh, 4 August 2026
- 23. GPT-5.2 Pro: 46.0% (± 7.8) — gpt-5.2-pro-2025-12-11_xhigh, xhigh, 13 June 2026
- 24. Gemini 3.7 Flash: 36.6% (± 7.6) — gemini-3.7-flash_high, high, 14 August 2026
- 25. Grok 4.6: 31.7% (± 7.4) — grok-4.6_xhigh, xhigh, 14 August 2026
- 25. Claude Opus 4.7: 31.7% (± 7.4) — claude-opus-4-7_max, max, 10 June 2026
- 27. GPT-5.2: 31.7% (± 7.3) — gpt-5.2-2025-12-11_xhigh, xhigh, 11 June 2026
- 28. Claude Sonnet 5: 29.3% (± 7.2) — claude-sonnet-5_max, max, 30 June 2026
- 29. DeepSeek-V4.1-Flash: 26.8% (± 7.0) — deepseek-v4.1-flash_max, max, 8 October 2026
- 29. Gemini 3.1 Pro: 26.8% (± 7.0) — gemini-3.1-pro-preview, 11 June 2026
- 29. Claude Opus 4.6: 26.8% (± 7.0) — claude-opus-4-6_max, max, 11 June 2026
- 29. Gemini 3.5 Flash: 26.8% (± 7.0) — gemini-3.5-flash_high, high, 10 June 2026
- 33. Grok 4.5: 24.4% (± 6.8) — grok-4.5_high, high, 9 July 2026
- 34. Gemini 3.8 Flash: 22.0% (± 6.5) — gemini-3.8-flash_high, high, 2 September 2026
- 34. Gemini 3.6 Flash: 22.0% (± 6.5) — gemini-3.6-flash_high, high, 2 August 2026
- 34. GPT-5: 22.0% (± 6.5) — gpt-5-2025-08-07_high, high, 11 June 2026
- 37. GPT-5 Pro: 19.5% (± 6.3) — gpt-5-pro-2025-10-06_high, high, 12 June 2026
- 38. Grok 4.7: 17.1% (± 5.9) — grok-4.7_xhigh, xhigh, 22 September 2026
- 38. Grok 4.20: 17.1% (± 5.9) — grok-4.20-0309-reasoning, 13 July 2026
- 38. Gemini 3 Flash: 17.1% (± 5.9) — gemini-3-flash-preview, 11 June 2026
- 41. GPT-5 Mini: 12.2% (± 5.2) — gpt-5-mini-2025-08-07_high, high, 12 June 2026
- 41. GPT-5.4 nano: 12.2% (± 5.2) — gpt-5.4-nano-2026-03-17_high, high, 12 June 2026
- 43. GPT-5.4 Mini: 9.8% (± 4.7) — gpt-5.4-mini-2026-03-17_xhigh, xhigh, 12 June 2026
- 44. o4-mini: 4.9% (± 3.4) — o4-mini-2025-04-16_high, high, 11 June 2026
- 44. Claude Opus 4.5: 4.9% (± 3.4) — claude-opus-4-5-20251101_32K, 32K, 11 June 2026
- 46. DeepSeek V4 Pro: 2.4% (± 2.4) — deepseek-v4-pro_max, max, 17 June 2026
- 46. Claude Opus 4.1: 2.4% (± 2.4) — claude-opus-4-1-20250805_32K, 32K, 11 June 2026
- 46. Claude Sonnet 4.5: 2.4% (± 2.4) — claude-sonnet-4-5-20250929_32K, 32K, 11 June 2026
- 46. GPT-5 nano: 2.4% (± 2.4) — gpt-5-nano-2025-08-07_high, high, 12 June 2026
- 50. Gemini 3.5 Flash-Lite: 0.0% (± 0.0) — gemini-3.5-flash-lite_high, high, 2 August 2026
- 50. o3-mini: 0.0% (± 0.0) — o3-mini-2025-01-31_high, high, 11 June 2026
- 50. Gemini 2.5 Pro: 0.0% (± 0.0) — gemini-2.5-pro, 11 June 2026
Epoch AI: the benchmark and its data · CC BY 4.0