Skip to content

Benchmarks

Sources
8
Entries tracked
22,310
Last sync
20 min
All sources syncing normally
  • Arena · Text7 hours ago
  • Arena · WebDev7 hours ago
  • Arena · Vision7 hours ago
  • Arena · Agent7 hours ago
  • GPQA Diamond21 minutes ago
  • SWE-bench21 minutes ago
  • Terminal-Bench20 minutes ago
  • ARC-AGI-221 minutes ago
FrontierMath ×111 of 69 models
ModelOrg± CISwingAdd to compare
1gpt-5.5-pro-pre-releaseOpenAI——52.4%49.5%–55.3%#23–60
2gpt-5.5-pre-releaseOpenAI——51.7%48.8%–54.6%#22–36
3gpt-5.4-pro-2026-03-05OpenAIMar 2026$30 / $18050.0%47.1%–52.9%#20–171
4gpt-5.4-2026-03-05OpenAISep 2026$2.50 / $1547.6%44.7%–50.5%#12–29
8gpt-5.2-2025-12-11OpenAIDec 2025$1.75 / $1440.7%37.8%–43.6%#34–92
16gpt-5-2025-08-07OpenAIAug 2025$1.25 / $1032.4%29.7%–35.2%#64–149
18gpt-5.1-2025-11-13OpenAINov 2025$1.25 / $1031.0%28.3%–33.8%#45–242
20gpt-5.4-mini-2026-03-17OpenAIOct 2026$0.75 / $4.5028.3%26.2%–30.3%#75–180
22gpt-5-mini-2025-08-07OpenAIAug 2025$0.25 / $227.2%24.6%–29.9%#126–343
24gpt-5.4-nano-2026-03-17OpenAIOct 2026$0.20 / $1.2525.9%23.3%–28.4%#95–222
41gpt-5-nano-2025-08-07OpenAIAug 2025$0.05 / $0.408.3%6.7%–9.9%#132–399

Showing FrontierMath exactly as the source published it — its own score and confidence interval, no blending. Select a second axis to build a weighted composite.

The few things in AI and quantum that actually mattered this week — honestly summarised.

One email a week. Unsubscribe in one click. No tracking, no sharing.