OpenAI-compatible model checks
Start with the model ID your SDK, Cursor, Codex CLI or gateway will actually send, then compare official API docs, OpenAI-compatible routes, global aggregators, Chinese model platforms and third-party relays. This page is for model ID and route verification, not live pricing or uptime claims.
Gateway model lists are useful only when they expose the exact model ID, provider route, context window and token pricing. Treat family names such as Claude Sonnet or GPT/Codex as starting points, then verify the concrete model ID your tool will send.
For OpenAI SDK-compatible tools, a model page is useful only when it helps you verify the exact model ID, `/v1/models` discovery behavior, supported endpoints, provider route and token pricing basis before a test request.
Recent queries for this page are less about a generic model list and more about exact docs, route shape and model IDs. Use these checks before choosing a gateway or SDK route.
High-capability Claude family used for complex coding, reasoning, and long-context work.
Balanced Claude family for daily coding assistance, agents, and editor workflows.
OpenAI GPT and Codex-compatible workflows for CLI tools, SDKs, and coding assistants.
Google model family for long-context, multimodal, and cost-sensitive agent workloads.
Meta's open-weight Llama model family, widely used across self-hosted, cloud, and OpenAI-compatible gateway workflows.
xAI's Grok model family for reasoning, conversational AI, tool use, and coding workflows with an OpenAI-compatible API.
Amazon's Nova model family is available through Amazon Bedrock for text, image, and video generation workflows.
Cohere's Command family focuses on enterprise chat, retrieval-augmented generation, multilingual tasks, embeddings, and reranking.
Microsoft's Phi small language model family is optimized for efficient reasoning, on-device scenarios, and Azure AI Foundry deployment.
AI21 Labs' Jamba family uses hybrid architectures for long-context enterprise text generation and document workflows.
Perplexity's Sonar API focuses on search-grounded chat, online answers, and citation-backed responses.
Mistral AI offers open-weight and hosted frontier models for multilingual, coding, agent, and enterprise workloads.
Cost-efficient reasoning and coding models with OpenAI and Anthropic compatible APIs. Strong open-weight ecosystem and agent tooling support.
Alibaba model family with strong Chinese, coding, multimodal and open-model ecosystem coverage. Offers OpenAI-compatible endpoints.
Zhipu AI's GLM family with strong reasoning, multimodal vision, image generation and OCR capabilities. Supports OpenAI and Claude-compatible APIs.
Moonshot AI's Kimi family, known for extremely long context windows (128K+), strong Chinese reasoning, and OpenAI-compatible APIs.
Full-modal model family covering text, video, voice, image and music generation. Strong in role-play and multimedia content creation.
Baichuan model family with strong Chinese-language capabilities, domain-specific fine-tuning, and enterprise-focused deployment options.
01.AI's Yi model family, optimized for bilingual (Chinese-English) reasoning, coding, and long-context understanding.
Baidu's ERNIE family, one of China's earliest and most mature LLM platforms, with strong Chinese understanding, multimodal capabilities, and deep Baidu ecosystem integration.
ByteDance's Doubao model family on the Volcano Engine Ark platform. Strong in Chinese conversation, function calling, and cost-efficient inference at scale.
iFlytek's Spark model family, built on decades of speech and NLP expertise. Strong in voice interaction, education, and vertical industry applications.
Tencent's Hunyuan model family, deeply integrated with WeChat and Tencent Cloud ecosystem. Strong in Chinese conversation, content creation, and enterprise workflows.
StepFun's Step model family, focused on multimodal understanding and generation with strong math and reasoning capabilities.
Kunlun Tech's Skywork model family with open-weight releases and a focus on Chinese-English bilingual capabilities and research accessibility.
SenseTime's SenseNova model family, built on computer vision expertise with strong multimodal understanding, image/video analysis, and edge deployment capabilities.
ModelBest's MiniCPM family, focused on small-but-capable edge-deployable models with competitive performance at tiny scales.
Prices and model names change quickly. Use official pricing pages and provider docs as the source of truth before topping up.
No. A gateway model list shows what that gateway currently exposes. Always compare the concrete model ID, upstream provider, context window and pricing against official model documentation before using it.
Prices can differ because gateways route through different upstream providers, add or remove markup, support BYOK routes, or bill cache, long context and media tokens differently.
Save the model ID, provider route, pricing page, request timestamp, token counts, error behavior and billing screenshot. Those records help detect pricing changes or unexpected provider routing.