US vs China · scores as of 19 October 2026
US vs China: who’s winning the AI race?
The US leads: the best American model (Gemini 4 Argon, 1524) is 37 points ahead of the best Chinese model (Kimi K3, 1487) on LMArena, up from 22 a year ago. China came closest in January 2025, when DeepSeek-R1 was just 4 points behind.
The gap over time
| Date | Best US model | Best Chinese model | Gap |
|---|---|---|---|
| July 2023 | GPT-4 · 1288 | — | — |
| January 2024 | GPT-4 Turbo Preview · 1313 | DeepSeek LLM 67B · 1185 | 128 |
| July 2024 | GPT-4o · 1346 | DeepSeek-Coder-V2 · 1265 | 81 |
| January 2025 | o1 · 1402 | DeepSeek-R1 · 1398 | 4 |
| July 2025 | GPT-4.5 Preview · 1445 | Qwen3-235B-A22B · 1422 | 23 |
| January 2026 | Gemini 3 Pro · 1485 | Kimi K2.5 · 1450 | 35 |
| July 2026 | Claude Fable 5 · 1504 | Kimi K3 · 1487 | 17 |
The labs on each side, live
| Lab | Country | AI Race rank | Index |
|---|---|---|---|
| Meta | US | #1 | 94.3 |
| Anthropic | US | #2 | 91.6 |
| Google DeepMind | US | #3 | 89.7 |
| Alibaba Qwen | China | #4 | 80.6 |
| OpenAI | US | #5 | 80.4 |
| DeepSeek | China | #6 | 73.8 |
| xAI | US | #7 | 70.8 |
| Moonshot AI | China | #8 | 58.8 |
| Mistral AI | Europe | #9 | 56.5 |
| Apple | US | #10 | 50.1 |
Model scores: LMArena text leaderboard, each country’s best model released by that date (today’s scores, so every year shares one scale). Google DeepMind is counted as US (Alphabet). Europe’s main contender is Mistral.