Tracking verified inference token volume and US/China ecosystem adoption (Daily 24h & Weekly 7d)
| Rank | Model & Organization | 24h Tokens | Global Share | 24h Change | Official API |
|---|---|---|---|---|---|
| #1 | DeepSeek V4-Flash | 5.12 T In 3.3T / Out 1.8T | 15.1% | +14.5% | Official API → |
| #2 | Doubao Pro 128k | 3.85 T In 3.0T / Out 0.9T | 11.4% | +16.8% | Official API → |
| #3 | Claude 3.5 Sonnet | 3.15 T In 2.6T / Out 0.5T | 9.3% | -1.8% | Official API → |
| #4 | GPT-4o | 2.82 T In 2.3T / Out 0.6T | 8.3% | +3.2% | Official API → |
| #5 | Doubao Lite 128k | 2.40 T In 1.9T / Out 0.6T | 7.1% | +18.2% | Official API → |
| #6 | Qwen 2.5 72B Instruct | 2.18 T In 1.6T / Out 0.6T | 6.4% | +11.2% | Official API → |
| #7 | Llama 3.3 70B Instruct | 1.95 T In 1.4T / Out 0.6T | 5.8% | +8.9% | Official API → |
| #8 | DeepSeek V3 | 1.62 T In 1.1T / Out 0.5T | 4.8% | +6.7% | Official API → |
| #9 | Claude 3.5 Haiku | 1.10 T In 0.9T / Out 0.2T | 3.2% | +4.8% | Official API → |
| #10 | Gemini 2.0 Flash | 980.00 B In 0.8T / Out 0.2T | 2.9% | +15.6% | Official API → |
| #11 | MiniMax-01 | 890.00 B In 0.7T / Out 0.2T | 2.6% | +9.3% | Official API → |
| #12 | GLM-4-Plus | 820.00 B In 0.6T / Out 0.2T | 2.4% | +7.1% | Official API → |
| #13 | GPT-4o-mini | 740.00 B In 0.6T / Out 0.2T | 2.2% | +1.2% | Official API → |
| #14 | Ernie 4.5 Turbo | 680.00 B In 0.5T / Out 0.2T | 2% | +8.5% | Official API → |
| #15 | Grok 2 | 590.00 B In 0.5T / Out 0.1T | 1.7% | +5.5% | Official API → |
| #16 | Qwen 2.5 Coder 32B | 550.00 B In 0.4T / Out 0.1T | 1.6% | +13.8% | Official API → |
| #17 | DeepSeek R1 | 510.00 B In 0.3T / Out 0.2T | 1.5% | +16.4% | Official API → |
| #18 | Llama 3.1 405B | 450.00 B In 0.3T / Out 0.1T | 1.3% | +3.1% | Official API → |
| #19 | Hunyuan Turbo | 420.00 B In 0.3T / Out 0.1T | 1.2% | +8.4% | Official API → |
| #20 | Mistral Large 2 | 400.00 B In 0.3T / Out 0.1T | 1.2% | +2.7% | Official API → |
| #21 | Claude 3 Opus | 330.00 B In 0.3T / Out 0.1T | 1% | -4.3% | Official API → |
| #22 | Yi-Lightning | 310.00 B In 0.2T / Out 0.1T | 0.9% | +5.9% | Official API → |
| #23 | Kimi K1.5 | 290.00 B In 0.2T / Out 0.1T | 0.9% | +7.8% | Official API → |
| #24 | Xiaomi MiMo 2.5 | 260.00 B In 0.2T / Out 0.1T | 0.8% | +12.5% | Official API → |
| #25 | Gemini 1.5 Pro | 240.00 B In 0.2T / Out 0.1T | 0.7% | +2.1% | Official API → |
| #26 | Baichuan 4 | 210.00 B In 0.2T / Out 0.1T | 0.6% | +4.2% | Official API → |
| #27 | o1-preview | 190.00 B In 0.1T / Out 0.1T | 0.6% | +6.1% | Official API → |
| #28 | Step-2 | 180.00 B In 0.1T / Out 0.0T | 0.5% | +5.3% | Official API → |
| #29 | DeepSeek Coder V2 | 170.00 B In 0.1T / Out 0.0T | 0.5% | +3.8% | Official API → |
| #30 | Qwen 2.5 7B | 150.00 B In 0.1T / Out 0.0T | 0.4% | +9.7% | Official API → |
[Methodology Disclosure] A single overseas gateway (e.g. OpenRouter) only reflects indie developers and overseas application pipelines, creating blind spots for China domestic enterprise usage (Alibaba DashScope, Baidu Qianfan, VolcEngine Doubao, SiliconFlow). To provide an objective benchmark, AIdaily adopts the Dual-Track Consensus Methodology (Consensus v1): 1. US and international models are benchmarked against OpenRouter global verified telemetry; 2. Chinese models are calibrated by blending SiliconFlow high-concurrency developer shares, ModelScope ecosystem activity, and verified official cloud milestones (such as VolcEngine Doubao trillion-token daily scale); 3. Country is assigned by laboratory headquarters. Data is refreshed daily at 17:30 PT and published atomically.