Baidu ERNIE Pricing 2026: What Does China's Other AI Giant Actually Charge?
Direct answer
Is ERNIE still economically competitive in China's price war? Baidu's official Qianfan billing documentation (cloud.baidu.com/doc/qianfan/s/wmh4sv6ya), updated September 16, 2026, lists ERNIE 5.1, Baidu's current flagship, at ¥4.00/¥18.00 per million tokens (input/output, for requests with ≤32K input tokens; ¥6.00/¥22.00 for the 32K–128K tier). ERNIE 5.0, the model 5.1 replaced, is actually more expensive — ¥6.00/¥24.00 at the same ≤32K tier — so Baidu's newer flagship costs less to run than its predecessor. Public materials cited here do not establish whether ERNIE 5.1 is purchasable on an international (non-mainland) Qianfan account.
Consumer product
ERNIE Bot (Baidu's consumer chat app) is reported free for individual users. Public materials do not clearly establish current subscription tiers or international availability.
API / Qianfan pricing, from Baidu's official billing page (cloud.baidu.com/doc/qianfan/s/wmh4sv6ya), updated by Baidu September 16, 2026. Original prices are published per thousand tokens in RMB; normalized here to per-million for comparability (×1,000), with an approximate USD conversion at an assumed ¥7.2 = $1 exchange rate — a stated assumption, not an official Baidu USD rate:
| Model | Input (¥/M → $) | Output (¥/M → $) | Tier |
|---|---|---|---|
| ERNIE 5.1 (current flagship) | ¥4.00 → $0.56 | ¥18.00 → $2.50 | ≤32K input |
| ERNIE 5.1 | ¥6.00 → $0.83 | ¥22.00 → $3.06 | 32K–128K input |
| ERNIE 5.0 (previous flagship) | ¥6.00 → $0.83 | ¥24.00 → $3.33 | ≤32K input |
| ERNIE 5.0 | ¥10.00 → $1.39 | ¥40.00 → $5.56 | 32K–128K input |
| ERNIE 4.5 Turbo (all current dated variants) | ¥0.80 → $0.11 | ¥3.20 → $0.44 | cache-hit input: ¥0.20 → $0.03 |
| ERNIE X1.1 Preview (reasoning) | ¥1.00 → $0.14 | ¥4.00 → $0.56 | Marked "即将下线" (retiring soon) directly on Baidu's own page |
Web search / grounding fee, same page: triggering search-enhancement costs ¥0.004 per use (≈$0.00056), plus the retrieved content's own token cost as ordinary input. Prepaid packages exist too: 10,000 searches for ¥38 (5% off ¥40 list), 50,000 for ¥190 (5% off ¥200).
A genuinely useful bonus finding from the same official page: Qianfan also resells several non-Baidu models at its own markup — GLM-5.3 is listed at ¥8.00/¥28.00 per million tokens on Qianfan, well above Z.ai's own direct rate of ¥2.10/¥8.40 (discounted) covered in the Z.ai article. If you're already inside the Qianfan ecosystem for ERNIE, check whether routing other-vendor models through Qianfan is actually cheaper than going direct — in GLM-5.3's case here, it isn't.
On international availability: the cited pricing is Baidu's mainland Qianfan billing in RMB. A separate source says the international Qianfan catalog lists ERNIE 5.0 but not 5.1, so the RMB rate does not establish that international accounts can purchase ERNIE 5.1. If you need it for a non-China deployment, verify availability on your own account before budgeting around the rate above.
Real workload calculations
CALCULATION, using ERNIE 5.1's confirmed ≤32K-tier rate (¥4.00/¥18.00 per million), shown in RMB first with the same ¥7.2=$1 stated conversion:
| Workload | Formula (RMB) | Cost (¥) | Cost (≈$) |
|---|---|---|---|
| A — Everyday (1M in, 250K out) | (1×¥4.00) + (0.25×¥18.00) | ¥8.50 | ≈$1.18 |
| B — Coding/reasoning (10M in, 3M out) | (10×¥4.00) + (3×¥18.00) | ¥94.00 | ≈$13.06 |
| C — High volume (400M in, 100M out/mo) | (400×¥4.00) + (100×¥18.00) | ¥3,400.00 | ≈$472.22 |
A brief note on the ≤32K tier: this is per individual request's input length, not total monthly volume — a realistic workload made of many separate requests, each well under 32K tokens of context, stays on this cheaper tier throughout even at high monthly volume, unless individual calls themselves regularly exceed 32K input tokens.
D — Provider-specific: the retiring reasoning model. ERNIE X1.1 Preview is explicitly marked for retirement on Baidu's own current billing page — a real, dated migration signal. Anyone routing production traffic to ERNIE-X1.1-Preview should plan a migration to ERNIE 5.0/5.1's own reasoning mode (both support "深度思考"/deep-thinking, per the same official page) before Baidu removes it.
Comparison with Qwen, DeepSeek, Doubao
A genuinely useful, primary-sourced comparison point: the same Baidu billing page also lists DeepSeek-V4.1-Flash as directly resold through Qianfan, at ¥2.00/¥8.00 per million (peak hours, 8:00–22:00) and ¥1.00/¥4.00 per million (off-peak, 22:00–8:00) — both confirmed on the same official page. At its peak rate, DeepSeek-V4.1-Flash on Qianfan (¥2/¥8) is meaningfully cheaper than ERNIE 5.1 (¥4/¥18) for the identical volume. This is a genuine, primary-sourced finding, not a cross-provider guess, since both rates come from the same Baidu document. Doubao's cheapest floor tier (covered elsewhere on NotCheapAI, ~$0.02/M input on Volcano Engine) remains cheaper still, though that comparison crosses providers and currencies and wasn't recomputed with the same rigor here.
BUILD & EARN scenario
MODELING ASSUMPTION, not a market benchmark: a Chinese-language enterprise document-processing product built on ERNIE 5.1 at Workload C-equivalent volume (¥3,400/month, ≈$472). At an illustrative 30 enterprise clients paying ¥1,000/month each (an assumed ¥30,000/month revenue), model cost is roughly 11% of revenue.
FAQ
Is ERNIE 5.1 available outside China? Genuinely unclear. We now have primary-confirmed mainland RMB pricing (¥4/¥18 per million, ≤32K tier) directly from Baidu's own billing documentation — but a separate source states the international Qianfan catalog still lists only ERNIE 5.0, not 5.1. The public materials cited here do not establish ERNIE 5.1 availability to international accounts.
Is ERNIE 5.0 or 5.1 cheaper? 5.1 is cheaper — a real, primary-confirmed finding: ¥4.00/¥18.00 per million (5.1, ≤32K) versus ¥6.00/¥24.00 (5.0, same tier). The newer flagship costs less to run, not more.
Is ERNIE X1 still active? The current variant, ERNIE X1.1 Preview, is explicitly marked for retirement on Baidu's own billing page as of this pass — budget for a migration to ERNIE 5.0/5.1's built-in deep-thinking mode instead of building new dependencies on it.
Is ERNIE still competitive in China's AI price war? On confirmed numbers, ERNIE 5.1 (¥4/¥18 per million) is priced above DeepSeek-V4.1-Flash's peak rate (¥2/¥8) when both are checked on the same Qianfan billing page — genuinely useful evidence that "competitive" depends on which specific model you're comparing against, not a blanket yes or no.