Skip to content

> Baidu (ERNIE)

ernie.baidu.com

Baidu ERNIE models are served through the Baidu Qianfan model service and Agent development platform on Baidu Cloud, alongside third-party models such as DeepSeek, Kimi and Qwen. From 24 October 2025, users who open Qianfan and accept the user agreement automatically receive new-user free token quota: 17 listed preset models each grant 1,000,000 tokens valid for 3 months, including ERNIE-4.5-Turbo-128K, ERNIE-4.5-Turbo-32K, ERNIE-4.5-Turbo-VL, ERNIE-X1-Turbo-32K, bge-large-en and bge-large-zh. Free quota offsets online inference of preset models only, not batch inference, and verified unit prices on the billing page include embeddings at 0.0005 CNY per thousand tokens, image generation from 0.05 CNY per image, and web search augmentation at 0.004 CNY per triggered search.

unknownbest-effort probe · checked never
Paid API API keyFreemiumMedium risk· conf 90%verified
Free credits
Not quantified
Official page states a new-user grant of 1,000,000 tokens on each of 17 preset models, valid 3 months (cloud.baidu.com Qianfan new-user free quota doc); no USD value is given and the ERNIE text-model unit-price table did not render on the official billing page, so no conversion was made (checked 2026-09-16).
Cheapest input /1M
Free
USD per 1M input tokens, best tracked model
Cheapest output /1M
Free
USD per 1M output tokens, best tracked model
Models tracked
5
Canonical models with normalized pricing
// source_analysisUprouter editorial analysis

// pros

  • First-time Qianfan users automatically receive 1,000,000 free tokens on each of 17 preset models, valid 3 months, granted after reading and accepting the user agreement; the list includes ERNIE-4.5-Turbo-128K, ERNIE-4.5-Turbo-32K, ERNIE-4.5-Turbo-VL and ERNIE-X1-Turbo-32K as well as DeepSeek, Kimi-K2 and Qwen3 models.
  • Embeddings and rerankers are cheap and uniformly priced: Embedding-V1, bge-large-zh, bge-large-en, tao-8k and Qwen3-Embedding models all cost 0.0005 CNY per thousand input tokens, with rerankers at 0.0005 to 0.0008 CNY per thousand tokens.
  • Image generation is metered per image at 1024x1024: musesteamer-air-image at 0.05 CNY and qwen-image at 0.25 CNY, with qwen-image-edit at 0.3 CNY per image.
  • OCR is available in both page and token units, with DeepSeek-OCR at 0.0003 CNY per thousand input tokens and 0.0012 CNY per thousand output tokens, and PaddleOCR-VL-0.9B and PP-StructureV3 at 0.09 CNY per page while a limited-time 50% discount applies.
  • Compute units for custom deployments can be bought postpaid and billed per minute with a 15-minute minimum charge (AI accelerator card type V at 0.32 CNY per minute list, 0.16 CNY promotional), or prepaid monthly at 9,000 CNY list, 4,500 CNY promotional.

// cons

  • Free tokens apply only to online inference of preset models and explicitly cannot be used for batch inference, and the grant is issued only on first activation of the platform with a 3-month expiry counted per model.
  • The text-generation unit-price table on the official billing page returned no values when fetched, so ERNIE 5.x and ERNIE 5.0 per-token prices could not be verified from public documentation and must be confirmed in the console.
  • Several published prices are limited-time promotions with a stated revert date, such as OCR returning from 0.09 CNY to 0.18 CNY per page after the 50% discount period and compute units having carried a promotion ending 30 June 2025.
  • Search augmentation is billed twice against one call (0.004 CNY per triggered search plus the search tokens added to the prompt) and continues to bill a delinquent account while a token pack or TPM prepayment remains, and all pricing is documented in Chinese with CNY per thousand tokens.

Use Qianfan when you specifically need ERNIE models or Chinese-language evaluation, and spend the 17 x 1,000,000 token grant on online inference within the 3-month window because it cannot be applied to batch jobs and does not carry over. Pair it with a domestic cloud budget in CNY, and treat the promotion-priced OCR and image endpoints as time-limited rather than as the steady-state rate.

// free_quota

Free tier present — value not quantified

This provider offers a free tier, but its terms don’t convert cleanly into a USD figure (e.g. request-count limits or capacity-dependent pools). We refuse to print a misleading zero here. See the note below and the official pricing page for the exact limits.

note: Official page states a new-user grant of 1,000,000 tokens on each of 17 preset models, valid 3 months (cloud.baidu.com Qianfan new-user free quota doc); no USD value is given and the ERNIE text-model unit-price table did not render on the official billing page, so no conversion was made (checked 2026-09-16).

// pricinglast verified Sep 16, 2026
PlanTypeMonthlyInput /1MOutput /1MNote
Pay-as-you-gopayg—FreeFreeCheapest tracked model rate (OpenRouter-normalized).
ERNIE 4.0 Turbopayg—FreeFreeRates vary significantly by model; see Qianfan dashboard for latest list.
pricing_last_verified: Sep 16, 2026official pricing page

Pricing normalized from public sources — always verify with the provider.

// model prices

ModelInput /1MOutput /1MContext
Baidu: ERNIE LiteFreeFree8k
Baidu: ERNIE SpeedFreeFree8k
Baidu: ERNIE 3.5$0.0000$0.00008k
Baidu: ERNIE 4.0 Turbo$0.0000$0.0000128k
ERNIE 4.5 VL 424B A47B $0.420$1.25123k
// faq
$ Does Baidu (ERNIE) have a free tier?

Yes — Baidu (ERNIE) offers a free tier. Always confirm current limits on Baidu (ERNIE)'s official pricing page — free tiers change without notice.

$ How much does Baidu (ERNIE) cost per 1M tokens?

The cheapest model we track at Baidu (ERNIE) is Free per 1M input tokens and Free per 1M output tokens. This is normalized from Baidu (ERNIE)'s published pricing — verify with the provider before purchasing, since prices change frequently.

$ Is Baidu (ERNIE) safe to use?

Uprouter rates Baidu (ERNIE) as medium risk. Medium to high risk for non-Chinese teams. The free grant is generous but narrow: it covers only online inference of the listed preset models, expires 3 months after issue, and cannot be applied to batch inference. Pricing transparency is the sharper issue, because the text-generation unit-price table did not render on the official billing page and multiple prices are limited-time promotions with published revert dates. Regional and regulatory considerations also apply to a Baidu Cloud account, and all billing units are CNY per thousand tokens with Chinese-language documentation only. Risk ratings are editorial, evidence-based, and never influenced by affiliate relationships.

// community0 reviews

Raw community sentiment, shown unblended. Votes never feed into objective data fields like price, risk, or status — and never into sort order.

No published reviews yet — be the first.

// more_informationprice history · live status · risk & confidence · code snippets

price_history

No recorded changes yet.

live_status

current: unknown

No probes recorded yet.

Uprouter runs its own probes; best-effort, not a guarantee. Probe results depend on our network location and cadence — for production, always use the provider’s own status page.

risk_&_confidence

Medium risk
90%

Medium to high risk for non-Chinese teams. The free grant is generous but narrow: it covers only online inference of the listed preset models, expires 3 months after issue, and cannot be applied to batch inference. Pricing transparency is the sharper issue, because the text-generation unit-price table did not render on the official billing page and multiple prices are limited-time promotions with published revert dates. Regional and regulatory considerations also apply to a Baidu Cloud account, and all billing units are CNY per thousand tokens with Chinese-language documentation only.

Confidence reflects how much verifiable evidence (official docs, probe history, community reports) backs this entry. Risk is our editorial assessment of operator reliability and terms — not financial advice.

code_snippets

# Baidu (ERNIE)'s own API — get a key from their dashboard
curl https://ernie.baidu.com/chat/completions \
  -H "Authorization: Bearer THEIR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "baidu/ernie-4.5-vl-424b-a47b",
    "messages": [{ "role": "user", "content": "Hello" }]
  }'

Direct to the provider — Uprouter receives nothing; verify their current pricing before purchasing.

// similar_providersscored by Paid API signals

Ranked by neutral similarity signals only — live status, free-tier shape and free-credit proximity. Commercial relationships never influence this list.