Quick take

Shanghai Telecom's token-compute package is worth tracking because it fits yangmao.ai's new direction: free credits, API pricing, and token-cost intelligence.

But it should not be marketed as a “free unlimited API”. A safer framing is: a low-cost telco AI token package test: is ¥1 for 250k quota points actually cheap?

Current details come from a 2026-05-16 CLS screenshot: Shanghai Telecom launched a token-compute package showing ¥1 = 250k quota points, mobile-bill payment, and access to multiple large models.

Do not treat quota points as tokens yet

The most important caveat: 250k quota points may not equal 250k real tokens.

AI platforms often apply different accounting rules:

  • input and output tokens may have different prices;
  • different models may burn quota at different multipliers;
  • long context, web access, or tool use may cost extra;
  • quota points may be an internal credit unit, not tokens.

    • So the real price must be measured from actual API calls, not inferred from the headline number alone.

    Fast price conversion

    If we use the most optimistic rough calculation:

  • ¥1 = 250k quota points;
  • 1M quota points = ¥4;
  • using roughly ¥7 = $1, that is about $0.57 per 1M quota points.

    • That would be competitive if quota points map closely to actual token usage. If a model burns multiple quota points per token, the effective cost goes up.

    Use the calculator here: /en/tools/telecom-token-calculator/

    Six things to verify

    1. Quota-point to token conversion

    Log input tokens, output tokens, and quota points consumed across at least three models.

    2. Supported models

    Check whether it supports DeepSeek, Qwen, Kimi, GLM, Hunyuan, or GPT/Claude-style third-party access.

    3. Individual API access

    Confirm whether individual users can obtain API keys, or whether the package is enterprise-only.

    4. Shanghai Telecom eligibility

    If only Shanghai Telecom numbers can activate it, conversion is narrower but the SEO/intelligence value remains high.

    5. Rate limits and expiry

    Low-cost packages often hide constraints: RPM, TPM, context length, quota expiry, and model-level limits.

    6. Renewal rules

    Because payment may use the phone bill, check whether renewal is enabled by default. For testing, buy only the minimum package and disable renewal if possible.

    Recommended test flow

  • Use a low-risk phone number or secondary number.
  • Buy only the minimum ¥1 package first.
  • Do not upload company code, contracts, private documents, or customer data.
  • Prepare three public prompts: short Q&A, long-text summary, and code explanation.
  • Log model, input tokens, output tokens, quota points used, latency, and errors.
  • Convert effective cost: actual spend / actual tokens * 1,000,000.

    • Comparison angle

    ChannelPricing signalPaymentKey riskBest use
    Shanghai Telecom Token Package¥1 for 250k quota points, conversion TBDMobile billquota/token mapping unknownlow-cost China API test
    SiliconFlowmodel-based token pricingChina-friendly paymentprice and limits can changeDeepSeek/Qwen/GLM access
    OpenRoutermulti-model USD pricingoverseas paymentChina access and billing frictionmulti-model fallback
    Alibaba Bailian / Volcano Arkcloud-provider pricingChina paymentquota, verification, model selectionenterprise and production use

    Who should test it

  • Developers in China looking for low-cost API access.
  • Indie hackers trying to reduce AI API cost.
  • Content creators tracking new AI API channels.
  • Technical decision makers evaluating telco AI packages as backup routes.

    • Who should avoid it

  • Anyone looking for a “free unlimited API”.
  • Production systems needing stable SLA today.
  • Workflows involving private data, contracts, customer data, or company code.
  • Users who cannot log token usage and quota-point burn.

yangmao.ai's view

This is more worth tracking than personal API relay platforms because it is a regulated telco channel. Its real value is not “freebie hunting”; it is a signal that AI tokens are becoming commoditized like mobile data packages.

The next tracking fields are: package entry, supported models, real quota burn, rate limits, expiry, renewal rules, and effective cost per million tokens. If real pricing is close to ¥4 per 1M tokens, this keyword deserves long-term coverage.