Atlas

Models

← All models

o1 Preview

OpenAI · 2024-09-12

o1-preview is OpenAI's September 2024 early-access release of its o1 reasoning model series, made available to users while the full o1 was still in development. It uses chain-of-thought reinforcement learning to spend more time "thinking" before responding, yielding substantially stronger performance on complex reasoning, coding, and math tasks compared to GPT-4o.

Benchmark scores

BenchmarkScore
Agentic Tool Use (Chat)55.1
Agentic Tool Use (Enterprise)66.4
AidanBench1875
Arabic1087
ARC-AGI-118.0
Arena Text — Business, Management & Financial Ops (No Style Control)1304
Arena Text — Chinese (No Style Control)1325
Arena Text — Creative Writing (No Style Control)1319
Arena Text — English (No Style Control)1383
Arena Text — Entertainment, Sports & Media (No Style Control)1332
Arena Text — Exclude Ties (No Style Control)1314
Arena Text — Exclude Ties (Style Controlled)1366
Arena Text — Expert (No Style Control)1337
Arena Text — French (No Style Control)1343
Arena Text — French (Style Controlled)1389
Arena Text — German (No Style Control)1309
Arena Text — Hard Prompts (No Style Control)1354
Arena Text — Hard Prompts English (No Style Control)1377
Arena Text — Instruction Following (No Style Control)1342
Arena Text — Japanese (No Style Control)1295
Arena Text — Korean (No Style Control)1291
Arena Text — Legal & Government (No Style Control)1346
Arena Text — Life, Physical & Social Science (No Style Control)1349
Arena Text — Longer Query (No Style Control)1344
Arena Text — Medicine & Healthcare (No Style Control)1306
Arena Text — Multi-turn (No Style Control)1369
Arena Text — Non-English (No Style Control)1315
Arena Text — Non-English (Style Controlled)1359
Arena Text — Occupational Mathematical (No Style Control)1380
Arena Text — Russian (No Style Control)1315
Loading Atlas data…