AI Proxy Providers · Cheap OpenAI / Claude API
Compare AI proxy providers offering OpenAI API proxy, Claude API proxy, and OpenAI-compatible services by pricing rules, supported models, payment methods, and selection guidance.
Anthropic / OpenAI / SpaceXAI / DeepSeek / Moonshot
GPT and Codex models start at 5% of official pricing, and Claude models at 18%. Input, output, and cache rates are public per 1M tokens and fluctuate with network load.
Anthropic / OpenAI / Google
GPT models start at about 5% of official pricing, Claude at about 21%, and Gemini at about 14%. Usage is billed per 1M tokens with no monthly fee; prices follow the currently published stable groups. The minimum top-up is $1 and balances do not expire.
Anthropic / OpenAI / SpaceXAI / Moonshot
GPT models start at about 5.1% of official pricing, Claude models at about 5.9%, and Gemini models at about 24.5%. Models are billed by token or call; daily, weekly, and monthly plans reset periodically and unused quota expires. Top-ups start at ¥6.8.
Anthropic / OpenAI / Google / SpaceXAI / DeepSeek
GPT models start at 5.2% of official pricing, Claude models at 5.5%, and Gemini models at 17.6%. Models are billed per 1M tokens; top-ups start at ¥6.8, and seven one-time plans range from ¥6.8 to ¥680. The first six are valid for 30 days and the ¥680 plan for 90 days.
Anthropic / OpenAI / Google
Gemini models start at about 6.7% of official pricing, while Claude and GPT models start at 50%. Models are billed in USD per 1M tokens. PAYGO top-ups start at ¥20 with ¥1 credited as $1 and balances do not expire; monthly plans cost about ¥0.31-¥0.44 per $1 of periodic credit, reset daily, and do not roll over.
Anthropic / OpenAI / Google / SpaceXAI / DeepSeek
GPT models start at 7% of official pricing, Claude models at 20%, and Gemini models at 80%. Usage is billed per token in USD per 1M tokens, with no subscriptions or account tiers, and balances do not expire. Stripe top-ups start at $3 and USDT top-ups at $15.
Anthropic / OpenAI / Google / SpaceXAI / DeepSeek / Moonshot
GPT models start at about 7.1% of official pricing, Claude at about 10%, and Gemini at about 14.3%. Comparisons use the public fixed Codex Pro (Codex only) 0.5x, Claude lite 0.7x, and Gemini sale-test 1x groups, with the page converting official pricing at $1 = ¥7. Top-ups credit $1 of platform balance per ¥1, start at ¥50, and balances do not expire.
Anthropic / OpenAI / Google / SpaceXAI / DeepSeek / Moonshot
GPT and Codex models start at about 7.4% of official pricing, Claude models at about 11.7%, and Gemini models at 21%. Top-ups use USD with usage-based or per-request billing; comparisons use the codex group for GPT and Codex, claude_lite for Claude, and discounts for Gemini. The minimum top-up is $5, signup includes $0.10 credit, and credits do not expire.
Anthropic / OpenAI / Google / DeepSeek / Moonshot
GPT models start at about 8.2% of official pricing, Claude at about 18.9%, and Gemini at about 25.2%. Billing uses USD and tokens, with a $5 minimum top-up, no monthly fee or subscription, and non-expiring credits; model discounts adjust with upstream costs.
Anthropic / OpenAI
GPT models start at about 9.5% of official pricing, while Claude models start at about 24%. Public model pricing is billed in USD per 1M tokens, with a $10 minimum top-up; current Claude input rates range from about $0.24-$2.45 and output rates from $1.22-$12.23, while GPT input rates range from about $0.071-$0.48 and output rates from $0.43-$2.85. GPT Image 2 is billed by image size.
Anthropic / OpenAI / Google / DeepSeek / Moonshot
The public model catalog uses token-based billing: GPT models start at 10% of official pricing, Claude at 35%, and Gemini at 40%; top-ups start at $1 and credits do not expire.
Anthropic / OpenAI / Google / SpaceXAI / DeepSeek / Moonshot
Among currently comparable prices, GPT models start at 10% of official pricing, Claude at 95%, and Gemini at 90%. Usage is billed per 1M tokens with no monthly minimum; top-ups start at $5.
Start here
Start with price, then check pricing rules, model coverage, payment methods, and invoice support. Once the real cost is clear, it is easier to tell which provider is worth it.
Start with how each provider explains plans, credits, multipliers, and overage pricing. A cheap headline is not enough; record the rules first, then calculate your actual cost.
A proxy may map expensive models to cheaper ones. Verify the model name, protocol compatibility, and response shape before you connect it.
Alipay, WeChat Pay, credit cards, and USDT affect top-up convenience. Teams should also check long-term payment and reimbursement workflows.
For teams, client projects, or company reimbursement, invoice support can matter more than one-off price. The filter only shows providers where invoice support has been observed.
Guides
Run the smallest working setup before deciding whether to use an AI proxy provider long term.
Start with tool type, protocol compatibility, Base URL, API Key, and model name.
Review the basic use of ANTHROPIC_BASE_URL, ANTHROPIC_AUTH_TOKEN, and API keys.
Understand how OPENAI_BASE_URL, OPENAI_API_KEY, the Responses API, and Codex configuration relate.
FAQ
A Claude API proxy is usually a third-party service that provides Anthropic or Claude-compatible endpoints. Developers use its Base URL and API key to route Claude Code, backend, or automation requests to the target model.
OpenAI API proxy services usually provide OpenAI-compatible endpoints, separate API keys, prepaid accounts, and custom Base URLs. Differences include pricing rules, available models, stability, payment methods, invoice support, and compatibility.
Prioritize four things: clear pricing rules, support for your payment methods, invoice availability, and coverage for the models and protocols required by Claude, Codex, Cursor, or similar tools.
Not always. Providers may bill by multipliers, plans, daily credits, balances, or model-specific rules. First confirm the rules, models, and payment terms, then calculate your actual usage cost.
A typical setup starts with a cloud server, an open-source proxy project, a domain, certificates, proxy configuration, and account authorization. Then users connect with a Base URL and API key.
Read the deployment guideIt is a forwarding layer: users send requests to the proxy, and the proxy routes them through its accounts, models, and rules to upstream services. The outside API looks familiar, while forwarding and billing happen in the middle.
