免費 / 開源高度效率的推理模型且 API 價格極低,但提供的專門多模態代理工具較少。
Gemini 4 Carbon (Gemini Carbon 4) 每月有 4,339,429 次造訪,月增率 +12.7%,定價模式為免費增值。Stable Diffusion API 每月有 26,802 次造訪,成長率 -21.8%,定價模式為付費。
| 面向/指標 | ![]() | ![]() |
|---|---|---|
| 官方網域 | deepmind.google | stablediffusionapi.com |
| 定價模式 | 免費增值 | 付費 |
| 每月造訪次數 | 433.94萬 (4,339,429) | 2.68萬 (26,802) |
| 月流量變化量 | +48.98萬 (+12.7%) | -7490 (-21.8%) |
| 主要市場與市佔 | 🇺🇸 美國 (23%) | 🇺🇸 美國 (16.5%) |
| 每月直接造訪 | 89.91萬 | 8529 |
| 使用者評分 | 4.8 | 4.2 |
| 支援平台 | 網站, API | 網站 |

Google DeepMind 榜首級資安防禦與長週期智慧體工程前沿模型 — Gemini 4 Carbon (Gemini Carbon 4),繼 Gemini 4 Argon 後的巔峰之作
在權威的 Artificial Analysis Intelligence Index v5.0 評測中斬獲 59 分高居榜首,全面超越 Claude Opus 5.5(58 分)與 GPT-6 Astra(57 分),在 RiemannBench 上達到 80.0%。
在 SEC-bench Pro 漏洞發掘基準達 88.2%,ExploitGym 漏洞驗證達 54.8%,CWE-bench 根因修復達 71.0%;安全程式碼生成率為 Gemini 4 Argon 的 2 倍,Chrome 團隊驗證修補品質提升 10 倍。
專為持續數天的複雜程式碼重構、全端開發與多步除錯工作流設計,在 WebDev Arena 取得 1768 Elo(+89 Elo 躍升),深度整合 Antigravity 與 Vertex AI。
原生支援文字、程式碼、影像、音訊與影片的 1M Token 上下文視窗,在 256k–1M 超長上下文基準 GraphWalks 上創下 96.2% F1 的頂尖檢索與圖推理精準度。

一個 REST 端點串接 SD 1.5、SDXL、Flux 與上萬個社群模型
一個 POST 端點即可處理 SD 1.5、SDXL、Flux、ControlNet 與 LoRA 模型的文生圖、圖生圖與編輯任務,並為每個模型提供 cURL、Python、Node 範例。
數千個社群 checkpoint 與 LoRA,涵蓋動漫、寫實與風格化系列,皆可查看參數與範例程式碼。
請求會回傳任務或 track id,耗時生成可透過輪詢或推送至自有 webhook 端點取得結果,不必阻塞客戶端。
Dreambooth 式端點可用自有圖片訓練專屬概念與風格,訓練出的模型 id 可直接用於生成呼叫。
免費 / 開源高度效率的推理模型且 API 價格極低,但提供的專門多模態代理工具較少。

免費 / 開源具備強大多語言支援的強大開源權重多模態模型,但需要自行託管或使用第三方推論。

專項更強Zhipu AI's robust GLM language and multimodal APIs, but with less global developer ecosystem support.
專項更強Zhipu AI's robust GLM language and multimodal APIs, but with less global developer ecosystem support.
專項更強MiniMax Image-01 is MiniMax's open-weight image generation model, available through the MiniMax API and MiniMax Design. It turns text prompts into high-resolution visuals and powers image workflows inside MiniMax's broader model family.
專項更強MiniMax Image-01 is MiniMax's open-weight image generation model, available through the MiniMax API and MiniMax Design. It turns text prompts into high-resolution visuals and powers image workflows inside MiniMax's broader model family.

免費 / 開源Lightweight open-weights models from Google for local deployment, but less capable than frontier closed models.
免費 / 開源Lightweight open-weights models from Google for local deployment, but less capable than frontier closed models.

一個 API 串接 1500+ 可上線的生成式媒體模型

免費 / 開源Together AI is the AI Native Cloud — a full-stack platform for running, fine-tuning and training open-source models. It spans serverless inference APIs, dedicated endpoints, GPU clusters, post-training and storage, all backed by in-house systems research.
免費 / 開源Together AI is the AI Native Cloud — a full-stack platform for running, fine-tuning and training open-source models. It spans serverless inference APIs, dedicated endpoints, GPU clusters, post-training and storage, all backed by in-house systems research.
專項更強Comprehensive multimodal foundation model family from Tencent, but primarily targeted at Asian enterprise markets.

免費 / 開源Replicate is a cloud API for running, fine-tuning, and deploying machine learning models. Spin up open-source and hosted models — from image, video, and audio generation to speech synthesis, captioning, and LLM inference — with one line of code, or package your own model with Cog and put it behind a production-ready API.
免費 / 開源Replicate is a cloud API for running, fine-tuning, and deploying machine learning models. Spin up open-source and hosted models — from image, video, and audio generation to speech synthesis, captioning, and LLM inference — with one line of code, or package your own model with Cog and put it behind a production-ready API.
一組相容 OpenAI 的 API,串接 500+ 圖像、影片、音訊與大模型