147API / 147AI Review: Pricing & Comparison

Enterprise AI API procurement: two common blockers

For many companies, choosing an AI API relay runs into two blockers that can’t be sidestepped:

1. Payment method: finance departments require corporate bank transfers — personal Alipay/WeChat Pay won’t do
2. Invoicing: a VAT invoice is needed for the books

Most domestic relays are positioned for individual developers — they can’t issue invoices or accept corporate transfers. 147API (147ai.com, operated by Shenzhen Open Future Intelligence Co., Ltd.) explicitly supports both — for enterprise procurement, this is the line between “usable” and “not usable,” not a minor feature.

Other basics: full coverage of mainstream models (Claude/GPT/Gemini/Grok), an open-source gateway new-api-dev maintained on GitHub, roughly a 50% discount on the newest flagship models, mainland direct connect, and a minimum top-up of about $2 (¥14).

Pricing in detail

147API uses RMB settlement and pay-as-you-go billing, with a minimum top-up of $2 (about ¥14) — a low bar for individual developers wanting to try the newest flagship models. The platform claims to have optimized multimodal API call pricing down to roughly 50% of official rates.

Per reference data from an industry comparison article (March 2026):

Model	Input price (/M tokens)	Output price (/M tokens)	Direct connect
Claude Fable 5	~$5 (official $10)	~$25 (official $50)	✓
Claude Opus 4.8	~$2.5 (official $5)	~$12.5 (official $25)	✓
Claude Sonnet 4.6	~$1.5 (official $3)	~$7.5 (official $15)	✓
GPT-5.5	~$15 (official $30)	~$30 (official $60)	✓
Gemini series	See official site	See official site	✓
Llama 4	See official site	See official site	✓

Note: the price ranges above come from 147API pricing cited in a third-party comparison article — actual pricing follows the 147AI website, so be sure to verify the latest rates before topping up.

Worth noting: 147API maintains a roughly 50% discount even on the newest flagship models (GPT-5.4, Claude 4.6) — a fairly competitive discount level among peers (some platforms shrink the discount on the newest models). RMB settlement and a $2 minimum top-up also make it relatively friendly to budget-constrained individual developers.

Model coverage and hands-on testing

147API’s core positioning is “full mainstream-model coverage + a unified multimodal gateway”:

• Anthropic: Claude Fable 5, Claude Opus 4.8, Claude Sonnet 4.6, Claude Haiku 4.5
• OpenAI: GPT-5.5, GPT-4o, the o1/o3 series
• Google: the Gemini 3.5/2.5 series
• Meta: the Llama 4 series (including multimodal)
• Mistral: Mistral Large 3
• xAI: Grok 4.3
• Multimodal: a unified gateway for image generation, image understanding, audio transcription, and other multimodal calls

147API has open-sourced the new-api-dev project on GitHub (an LLM gateway framework) — making its tech stack public helps build platform credibility and suggests some genuine in-house technical development underneath, rather than a pure resale model.

A March 2026 industry comparison article (“Which of the Big Five Platforms Is Worth Choosing”) named 147API its go-to recommendation, praising it for “extremely fast rollout of new models, pricing held around 50% of official rates, first-tier response speed, and support for small-scale trial-and-error” — but note that this review originates from 147API’s own corporate blog (cnblogs.com/147api), so it carries some inherent bias.

Using it day to day

Signup: go to 147ai.com and register with an email address; RMB settlement is supported. A minimum top-up of $2 (about ¥14) gets you started, and the whole process takes about 5 minutes.

Payment: RMB settlement is supported (check the official site for specific payment methods), and it supports corporate bank transfers — a significant advantage for enterprise users who need invoices, since most small relays don’t offer corporate transfers or invoicing.

Getting your API key: generate one in the console; the endpoint is https://147ai.com/v1, fully compatible with the OpenAI SDK. Tools like Claude Code and Cursor need only swap in the base_url to connect.

Multimodal gateway: the open-source new-api-dev framework provides a unified multimodal interface — text, image, and audio calls all go through the same API address, simplifying integration for multimodal applications.

Invoicing and corporate transfers: compared with most individual-oriented relays, 147API explicitly supports corporate payment and invoicing — suited to commercial projects with financial-compliance requirements.

Pricing transparency: publicly available information on the site is limited — detailed pricing and SLA terms require contacting the official team directly. This is a clear shortcoming compared to 302.AI (which has a model marketplace with visible pricing).

Stability and latency

147API supports mainland direct connect, no proxy needed. Per industry comparison data (2026, some of it self-reported by the vendor):

• Evening-peak TTFT (time to first token): consistently under 300ms (per data cited in an industry review)
• Friday 9pm load test: average latency 320ms, no packet loss (evaluate the credibility of this specific test source for yourself)

These figures come from an affiliated review article and haven’t been comprehensively verified by an independent third party. That said, compared to competitors whose numbers are purely vendor self-reported, 147API’s ~300ms latency range shows up fairly consistently across third-party review citations, giving it slightly more reference value.

No detailed official SLA terms or compensation mechanism were found in public materials — for enterprise use, we’d recommend confirming these in writing beforehand.

Who it fits

• Enterprise developers (needing corporate transfers/invoices): 147API’s corporate payment and invoicing support is its differentiating advantage for enterprise users, suited to commercial projects with financial-compliance needs.
• Individual developers trying new models cheaply: a $2 minimum top-up and roughly 50% off the newest flagships suits developers who want to try Claude Fable 5 / GPT-5.5 at the lowest possible cost.
• Multimodal application developers: the unified multimodal gateway simplifies integration work across text, image, and audio.
• Claude Code use cases: ~300ms latency suits AI coding assistants, with friendly mainland direct connect.

Not a good fit: use cases that need to frequently check exact pricing or need an extremely low top-up threshold (like starting at ¥10) for rapid trial-and-error; or highly compliance-focused enterprises with strict requirements for a published SLA.

Head-to-head comparisons

• vs. 4SAPI: both are positioned as enterprise-grade aggregators with similar pricing strategy (roughly 50% off). Industry comparison data shows 4SAPI’s TTFT at about 520ms vs. 147API’s ~300ms — if latency is the deciding factor, 147API has a slight edge, though 4SAPI has more marketing exposure and slightly higher brand recognition. Neither platform’s latency figures come from a fully independent third party, so testing it yourself is recommended.
• vs. PoloAPI: PoloAPI offers usage statistics and multi-project call auditing, while 147API’s edge is corporate transfer/invoicing support suited to enterprises with financial needs. If your core need is “save money + pay via corporate transfer,” go with 147API; if it’s “team-level usage management,” go with PoloAPI.

Bottom line

147API is one of the go-to platforms named across multiple industry comparison articles, with clear advantages in interface compatibility (OpenAI-compatible plus the open-source new-api-dev framework), discounts on the newest flagship models (roughly 50% off), and corporate transfer support for enterprises. Its main shortcomings are limited pricing transparency on the official site (requiring direct inquiry) and the fact that most of its endorsing data comes from an affiliated blog, lacking fully independent third-party verification. For development teams that need corporate payment support, want roughly 50%-off access to the newest models, and are willing to contact the vendor directly to verify the details, 147API is worth serious consideration.

If you need to sign a DPA data-processing agreement (for finance, healthcare, or government-procurement use cases), CloseAI offers a formal, signable DPA — a more compliance-oriented direction.

Information verified 2026-06-13. For specific pricing, SLA, and latency figures, readers should verify against the latest official site information themselves.

Related reviews

• DeepSeek: DeepSeek’s official API, with V4-series pricing that upended the market and strong reasoning ability
• AIFast.club: 572+ models with mainland direct connect, bulk discounts, a real-time status monitoring dashboard, unified management under a single key
• ChatFire: a low-price multimodal aggregator, exchange rate ¥0.5-1/USD, a mix of domestic and international models, includes image and video generation
• Bob API: an individual-developer-style platform, high-quality mainland direct connect, mainstream models, suited to personal projects