| # | Company | HQ | Score | vs #1 | Engine | Deliver | Accel | |
|---|---|---|---|---|---|---|---|---|
| 01 | 🇺🇸 | GoogleDeepMind, Google Cloud AI, Search AI, Gemini, AlphaFold, Veo, Waymo | US | 9.46 | — | 9.7 | 9.4 | 9.0 |
| 02 | 🇺🇸 | OpenAIGPT series, ChatGPT, DALL·E, Sora, Codex | US | 8.85 | -0.61 | 8.8 | 9.0 | 8.5 |
| 03 | 🇺🇸 | AnthropicClaude model family, Claude Code, Claude API | US | 8.38 | -1.08 | 8.0 | 8.4 | 9.1 |
| 04 | 🇺🇸 | MicrosoftCopilot (Office, GitHub, Windows), Azure AI, Phi models | US | 8.35 | -1.12 | 8.1 | 9.0 | 8.6 |
| 05 | 🇺🇸 | MetaLlama model family, FAIR research, AI-powered feeds & ads, Meta AI assistant | US | 8.07 | -1.39 | 8.3 | 8.2 | 7.4 |
| 06 | 🇨🇳 | AlibabaQwen model family, Tongyi Qianwen, Alibaba Cloud AI | CN | 7.87 | -1.59 | 8.1 | 7.8 | 7.4 |
| 07 | 🇨🇳 | ByteDanceDoubao models, Dreamina, Coze, TikTok recommendation AI | CN | 7.82 | -1.64 | 7.8 | 8.3 | 7.4 |
| 08 | 🇺🇸 | AmazonAWS Bedrock, Trainium/Inferentia chips, Nova models, Alexa+, Rufus | US | 7.77 | -1.69 | 7.6 | 8.3 | 8.0 |
| 09 | 🇨🇳 | BaiduERNIE 5.0, Ernie Bot, Baidu Cloud AI, custom AI chips | CN | 7.25 | -2.21 | 7.4 | 7.4 | 6.6 |
| 10 | 🇺🇸 | xAIGrok, Colossus supercomputer | US | 7.23 | -2.23 | 7.7 | 6.7 | 6.4 |
| 11 | 🇨🇳 | DeepSeekDeepSeek-V3, DeepSeek-R1 | CN | 7.21 | -2.25 | 7.5 | 6.8 | 6.5 |
| 12 | 🇫🇷 | Mistral AIMistral Large, Mixtral, Le Chat, La Plateforme | FR | 7.05 | -2.41 | 6.9 | 7.0 | 7.4 |
| 13 | 🇺🇸 | AppleApple Intelligence, on-device models, Siri | US | 6.80 | -2.66 | 6.3 | 7.8 | 7.5 |
| 14 | 🇨🇳 | Moonshot AIKimi K2.6, Kimi chatbot | CN | 6.50 | -2.96 | 6.5 | 6.4 | 6.3 |
| 15 | 🇨🇳 | Zhipu AIGLM-5, GLM-5.1, ChatGLM, Z.ai platform | CN | 6.24 | -3.22 | 6.2 | 6.2 | 6.2 |
Each company is independently scored across ten dimensions by a panel of frontier evaluator models. Scores are averaged, weighted by a structural profile that prizes long-horizon inputs, and published with the per-score justifications intact.
Read the methodology →