跨评分、稳定性、动量和市场份额对比AI服务商。排名基于 300 个追踪模型的综合评分,每小时更新。
| # | 提供商 | 模型 | 平均评分 | 最佳模型 | 最高排名 | 24小时平均 | 7天平均 | 稳定性% | 免费 |
|---|---|---|---|---|---|---|---|---|---|
| 1 | Anthropic | 14 | 76.4 | Claude Opus 4.6 (Fast)(90.4) | #4 | -0.1 | -0.2 | 100% | 0 |
| 2 | xAI | 11 | 72.0 | Grok 4.20(88.8) | #8 | -0.2 | -0.5 | 100% | 0 |
| 3 | OpenAI | 57 | 71.4 | GPT-5.4 Pro(91.9) | #1 | -0.3 | +2.4 | 98% | 2 |
| 4 | 23 | 68.4 | Gemini 3 Flash Preview(88.4) | #11 | 0 | 0 | 100% | 4 | |
| 5 | MiniMax | 8 | 68.3 | MiniMax M2.5 (free)(78.2) | #49 | -0.1 | +0.4 | 100% | 1 |
| 6 | Zhipu AI | 12 | 67.9 | GLM 5(78.0) | #51 | -0.3 | -0.4 | 100% | 1 |
| 7 | DeepSeek | 12 | 66.9 | R1 0528(79.4) | #40 | -0.1 | -0.3 | 100% | 0 |
| 8 | StepFun | 1 | 66.8 | Step 3.5 Flash(66.8) | #119 | -1.0 | 0 | 100% | 0 |
| 9 | Cursor | 2 | 65.7 | Composer 2(65.7) | #125 | -1.0 | -1.0 | 100% | 0 |
| 10 | Inception | 1 | 61.0 | Mercury 2(61.0) | #144 | -1.0 | -1.0 | 100% | 0 |
| 11 | Xiaomi | 5 | 60.3 | MiMo-V2.5-Pro(75.8) | #61 | 0 | -1.4 | 100% | 0 |
| 12 | Moonshot AI | 5 | 58.4 | Kimi K2.6(75.9) | #60 | -0.8 | -0.2 | 100% | 0 |
| 13 | Microsoft | 2 | 56.5 | Phi 4(60.2) | #149 | -1.0 | +87.0 | 50% | 0 |
| 14 | Meta | 9 | 55.5 | Llama 4 Maverick(67.1) | #114 | -0.8 | -0.1 | 100% | 2 |
| 15 | Allen AI | 1 | 54.9 | Olmo 3 32B Think(54.9) | #161 | -1.0 | 0 | 100% | 0 |
| 16 | Tencent | 2 | 54.5 | Hy3 preview(69.0) | #108 | +47.0 | +116.5 | 50% | 0 |
| 17 | Alibaba | 49 | 52.2 | Qwen3.5 397B A17B(79.6) | #39 | -0.4 | -0.9 | 100% | 2 |
| 18 | Cohere | 3 | 49.4 | Command A(50.8) | #170 | -0.7 | -0.7 | 100% | 0 |
| 19 | arcee-ai | 6 | 48.1 | Trinity Large Thinking(65.2) | #129 | -0.3 | -1.2 | 100% | 0 |
| 20 | Mistral AI | 18 | 46.1 | Mistral Large 3 2512(67.0) | #115 | -0.4 | +7.7 | 94% | 0 |
| 21 | Amazon | 4 | 45.1 | Nova 2 Lite(60.5) | #148 | -0.3 | +0.5 | 100% | 0 |
| 22 | Baidu | 7 | 42.8 | ERNIE 4.5 300B A47B (59.6) | #152 | -0.3 | +21.0 | 86% | 2 |
| 23 | NVIDIA | 9 | 42.3 | Llama 3.3 Nemotron Super 49B V1.5(60.6) | #146 | -0.2 | -2.7 | 100% | 5 |
| 24 | inclusionai | 3 | 40.0 | Ring-2.6-1T (free)(40.0) | #183 | -0.7 | +98.7 | 33% | 1 |
| 25 | IBM | 2 | 40.0 | Granite 4.1 8B(40.0) | #186 | -0.5 | -2.5 | 100% | 0 |
| 26 | poolside | 2 | 40.0 | Laguna XS.2 (free)(40.0) | #189 | -1.0 | -4.0 | 100% | 2 |
| 27 | ~anthropic | 3 | 40.0 | Anthropic Claude Haiku Latest(40.0) | #191 | -0.7 | -3.7 | 100% | 0 |
| 28 | ~openai | 2 | 40.0 | OpenAI GPT Mini Latest(40.0) | #192 | -1.0 | -4.0 | 100% | 0 |
| 29 | 2 | 40.0 | Google Gemini Pro Latest(40.0) | #193 | -1.0 | -4.0 | 100% | 0 | |
| 30 | ~moonshotai | 1 | 40.0 | MoonshotAI Kimi Latest(40.0) | #194 | -1.0 | -4.0 | 100% | 0 |
| 31 | Kuaishou | 1 | 40.0 | KAT-Coder-Pro V2(40.0) | #209 | 0 | -3.0 | 100% | 0 |
| 32 | rekaai | 2 | 40.0 | Reka Edge(40.0) | #210 | 0 | -0.5 | 100% | 0 |
| 33 | ByteDance | 5 | 40.0 | Seed-2.0-Lite(40.0) | #215 | 0 | -2.8 | 100% | 0 |
| 34 | Liquid AI | 3 | 40.0 | LFM2-24B-A2B(40.0) | #217 | 0 | -3.0 | 100% | 2 |
| 35 | aion-labs | 3 | 40.0 | Aion-2.0(40.0) | #218 | 0 | +1.0 | 100% | 0 |
| 36 | Upstage | 1 | 40.0 | Solar Pro 3(40.0) | #221 | 0 | -3.0 | 100% | 0 |
| 37 | Writer | 1 | 40.0 | Palmyra X5(40.0) | #222 | 0 | -3.0 | 100% | 0 |
| 38 | essentialai | 1 | 40.0 | Rnj 1 Instruct(40.0) | #232 | 0 | -3.0 | 100% | 0 |
| 39 | deepcogito | 1 | 40.0 | Cogito v2.1 671B(40.0) | #238 | 0 | -3.0 | 100% | 0 |
| 40 | Perplexity | 5 | 40.0 | Sonar Pro Search(40.0) | #240 | 0 | +1.8 | 100% | 0 |
| 41 | AI21 Labs | 1 | 40.0 | Jamba Large 1.7(40.0) | #261 | 0 | -2.0 | 100% | 0 |
拥有至少2个模型的服务商,按处于"稳定"状态的模型百分比排名。更高的稳定性意味着更一致的长期表现。
追踪模型在各服务商间的分布。显示每个服务商在排行榜总计 300 个模型中的份额。
| 提供商 | 模型 | 份额 | % |
|---|---|---|---|
| OpenAI | 57 | 19.0% | |
| Alibaba | 49 | 16.3% | |
| 23 | 7.7% | ||
| Mistral AI | 18 | 6.0% | |
| Anthropic | 14 | 4.7% | |
| Zhipu AI | 12 | 4.0% | |
| DeepSeek | 12 | 4.0% | |
| xAI | 11 | 3.7% | |
| Meta | 9 | 3.0% | |
| NVIDIA | 9 | 3.0% | |
| MiniMax | 8 | 2.7% | |
| Baidu | 7 | 2.3% | |
| arcee-ai | 6 | 2.0% | |
| Xiaomi | 5 | 1.7% | |
| Moonshot AI | 5 | 1.7% | |
| ByteDance | 5 | 1.7% | |
| Perplexity | 5 | 1.7% | |
| Amazon | 4 | 1.3% | |
| Cohere | 3 | 1.0% | |
| inclusionai | 3 | 1.0% | |
| ~anthropic | 3 | 1.0% | |
| Liquid AI | 3 | 1.0% | |
| aion-labs | 3 | 1.0% | |
| Cursor | 2 | 0.7% | |
| Microsoft | 2 | 0.7% | |
| Tencent | 2 | 0.7% | |
| IBM | 2 | 0.7% | |
| poolside | 2 | 0.7% | |
| ~openai | 2 | 0.7% | |
| 2 | 0.7% | ||
| rekaai | 2 | 0.7% | |
| StepFun | 1 | 0.3% | |
| Inception | 1 | 0.3% | |
| Allen AI | 1 | 0.3% | |
| ~moonshotai | 1 | 0.3% | |
| Kuaishou | 1 | 0.3% | |
| Upstage | 1 | 0.3% | |
| Writer | 1 | 0.3% | |
| essentialai | 1 | 0.3% | |
| deepcogito | 1 | 0.3% | |
| AI21 Labs | 1 | 0.3% |
Providers are compared across multiple dimensions: average composite score across all models, best individual model score, stability rate (percentage of models in a stable state), 24-hour and 7-day momentum, model count, and availability of free models. Data is derived from hourly-updated rankings of 290+ tracked models.
The stability percentage represents the fraction of a provider's models that are in a "stable" ranking state. A higher percentage means the provider's models consistently maintain their positions over time, suggesting reliable and predictable performance across their lineup.
Market concentration is shown as each provider's share of the total tracked models. The visual bar chart and percentage table reveal how models are distributed among providers, helping identify whether the market is dominated by a few large players or spread across many competitors.