Comparisons
kie.ai pricing: median 70% under list price
kie.ai pricing vs list prices for 31 AI text models, fetched 2026-09-24. Compare per-token rates, the bill at 1M calls, and 5 cases where kie.ai falls short.
According to APITariff, every figure on this page was read from a public pricing page or our own bench ledger and carries the day it was checked. Data as of 2026-09-24, across 1782 models. How it is collected.
According to APITariff, kie.ai pricing undercuts the list price on 31 of the 32 AI text models it shares with OpenRouter, with a median saving of 70% per 1M output tokens, fetched 2026-09-24. We fetched all 495 rows of the kie.ai pricing API and compared its 37 chat models with OpenRouter’s price for the same model, matched by exact name. 29 of the 31 cheaper models save 50% or more on output. At 1,000,000 calls of 1,000 input and 300 output tokens, Claude Opus 5 bills $5,000 through kie.ai against $12,500 at list, and GPT-5.6 Luna bills $156.80 against $560. The one match that is not cheaper on both sides is DeepSeek V4.1 Flash, where kie.ai charges $0.12 per 1M input tokens against $0.04 in our last clean OpenRouter row. kie.ai is a reseller. It is not Anthropic, OpenAI, Google or xAI, and it bills in credits worth $0.005 each.
Last updated 2026-09-24 · 31 models · 5 providers · sources: kie.ai pricing API, APITariff daily fetch of OpenRouter, docs.kie.ai.
kie.ai pricing vs list prices, model by model
OpenAI’s GPT-5.2 shows the widest gap, 75% under list on output, and GPT-5.6 Sol the narrowest, 16%. The table lists all 31 models where kie.ai is cheaper on both input and output, sorted by output saving.
| Model | Provider | kie.ai in / out | List in / out | Output saving | 1M calls via kie.ai | 1M calls at list |
|---|---|---|---|---|---|---|
| GPT-5.2 | OpenAI | 0.44 / 3.5 | 1.75 / 14 | 75% | $1,490.00 | $5,950.00 |
| GPT-5.5 | OpenAI | 1.4 / 8.4 | 5 / 30 | 72% | $3,920.00 | $14,000.00 |
| GPT-5.6 Luna | OpenAI | 0.056 / 0.336 | 0.2 / 1.2 | 72% | $156.80 | $560.00 |
| GPT-5.6 Terra | OpenAI | 0.56 / 3.36 | 2 / 12 | 72% | $1,568.00 | $5,600.00 |
| GPT-6 Astra | OpenAI | 2.8 / 14 | 10 / 50 | 72% | $7,000.00 | $25,000.00 |
| Claude Haiku 4.5 | Anthropic | 0.275 / 1.425 | 1 / 5 | 71.5% | $702.50 | $2,500.00 |
| Claude Sonnet 4.5 | Anthropic | 0.85 / 4.275 | 3 / 15 | 71.5% | $2,132.50 | $7,500.00 |
| Claude Sonnet 4.6 | Anthropic | 0.85 / 4.275 | 3 / 15 | 71.5% | $2,132.50 | $7,500.00 |
| Claude Opus 4.5 | Anthropic | 1.425 / 7.15 | 5 / 25 | 71.4% | $3,570.00 | $12,500.00 |
| Claude Opus 4.6 | Anthropic | 1.425 / 7.15 | 5 / 25 | 71.4% | $3,570.00 | $12,500.00 |
| Claude Opus 4.7 | Anthropic | 1.425 / 7.15 | 5 / 25 | 71.4% | $3,570.00 | $12,500.00 |
| Gemini 2.5 Flash | 0.09 / 0.75 | 0.3 / 2.5 | 70% | $315.00 | $1,050.00 | |
| Gemini 2.5 Pro | 0.38 / 3 | 1.25 / 10 | 70% | $1,280.00 | $4,250.00 | |
| Gemini 3.5 Flash | 0.45 / 2.7 | 1.5 / 9 | 70% | $1,260.00 | $4,200.00 | |
| Gemini 3.6 Flash | 0.225 / 1.125 | 0.75 / 3.75 | 70% | $562.50 | $1,875.00 | |
| Gemini 3.7 Flash | 0.225 / 1.125 | 0.75 / 3.75 | 70% | $562.50 | $1,875.00 | |
| Gemini 3.8 Flash | 0.225 / 1.125 | 0.75 / 3.75 | 70% | $562.50 | $1,875.00 | |
| GPT-5.4 | OpenAI | 0.7 / 5.6 | 2.5 / 15 | 62.7% | $2,380.00 | $7,000.00 |
| Claude Fable 5 | Anthropic | 4 / 20 | 10 / 50 | 60% | $10,000.00 | $25,000.00 |
| Claude Opus 4.8 | Anthropic | 2 / 10 | 5 / 25 | 60% | $5,000.00 | $12,500.00 |
| Claude Opus 5 | Anthropic | 2 / 10 | 5 / 25 | 60% | $5,000.00 | $12,500.00 |
| GPT-5.1-Codex | OpenAI | 0.5 / 4 | 1.25 / 10 | 60% | $1,700.00 | $4,250.00 |
| GPT-5.2-Codex | OpenAI | 0.7 / 5.6 | 1.75 / 14 | 60% | $2,380.00 | $5,950.00 |
| GPT-5.3-Codex | OpenAI | 0.7 / 5.6 | 1.75 / 14 | 60% | $2,380.00 | $5,950.00 |
| Grok 4.3 | xAI | 0.5 / 1 | 1.25 / 2.5 | 60% | $800.00 | $2,000.00 |
| Grok 4.5 | xAI | 0.8 / 2.4 | 2 / 6 | 60% | $1,520.00 | $3,800.00 |
| Grok 4.6 | xAI | 0.8 / 2.4 | 2 / 6 | 60% | $1,520.00 | $3,800.00 |
| Claude Sonnet 5 | Anthropic | 0.85 / 4.275 | 2 / 10 | 57.2% | $2,132.50 | $5,000.00 |
| Grok 4.7 | xAI | 0.8 / 2.4 | 1.6 / 4.8 | 50% | $1,520.00 | $3,040.00 |
| Kimi K3 | Moonshot AI | 2.4 / 12 | 3 / 15 | 20% | $6,000.00 | $7,500.00 |
| GPT-5.6 Sol | OpenAI | 1.4 / 8.4 | 2 / 10 | 16% | $3,920.00 | $5,000.00 |
Prices in USD per 1M tokens. kie.ai column: kie.ai pricing API, usdPrice field, read 2026-09-24. List column: OpenRouter, fetched 2026-09-24T07:24Z; for 10 of the 11 Anthropic and OpenAI rows we checked against the providers’ own pricing pages on 2026-09-24, it equals the provider’s standard rate, and GPT-5.6 Sol is the exception. 1M calls = 1,000 input + 300 output tokens per call. 31 of 31 cheaper matches shown.
Cost at three volumes
At 10,000 calls a month the saving on Claude Opus 5 is $75. At 1,000,000 calls it is $7,500. The rows below use the same 1,000 input and 300 output tokens per call.
| Model | 10k calls kie.ai / list | 100k calls kie.ai / list | 1M calls kie.ai / list |
|---|---|---|---|
| GPT-6 Astra | $70 / $250 | $700 / $2,500 | $7,000 / $25,000 |
| Claude Opus 5 | $50 / $125 | $500 / $1,250 | $5,000 / $12,500 |
| Claude Sonnet 5 | $21.33 / $50 | $213.25 / $500 | $2,132.50 / $5,000 |
| GPT-5.6 Luna | $1.57 / $5.60 | $15.68 / $56 | $156.80 / $560 |
Computed from the per-token rates in the table above, 2026-09-24. kie.ai allows 20 new requests per 10 seconds per account, about 5.2 million in a 30-day month; 1,000,000 calls spread evenly over 30 days is 0.39 per second against that limit of 2. For other token mixes, the token calculator takes any input and output count.
Where going direct wins, or kie.ai falls short
Five cases weaken the kie.ai price on the data read 2026-09-24: one unsettled price, two missing models, two small or partial savings.
- DeepSeek V4.1 Flash is unsettled: no two sources agree on its input price yet. kie.ai lists $0.12 input and $0.475 output. Our last clean OpenRouter row, fetched 2026-09-23, lists $0.04 and $0.64, so 1M calls cost $262.50 through kie.ai and $232 at that rate. On 2026-09-24 OpenRouter moved the input price to $0.14, and our engine held that 250% jump back for review. DeepSeek’s own pricing page lists $0.15 input off-peak and $0.30 at peak, so the DeepSeek V4.1 Flash record needs a second fetch before either side can claim it.
- Claude Opus 5.5 and Claude Fable 5.1 have no kie.ai price row. kie.ai lists the older Claude Opus 5 and Claude Fable 5 instead.
- GPT-5.6 Sol saves 16% on output. OpenRouter lists Sol at $2 and $10. OpenAI’s pricing page, read 2026-09-24, lists $4 and $20 as standard, prints $2 and $10 only in its Batch and Flex tables, and says: “GPT-5.6 Sol’s promotional pricing is available at least through November 21, 2026.” Against OpenAI’s standard $4 and $20, kie.ai’s $1.40 and $8.40 would save 65% and 58%. The GPT-5.6 Sol record tracks the rate.
- Kimi K3 saves 20%. kie.ai lists $2.40 and $12 against $3 and $15.
- Cached input is not compared here. kie.ai lists cached-input and cache-write rows for some models, such as $0.14 cached input for GPT-5.6 Sol. Workloads that reuse a long prompt should price the cached rate on both sides.
How does kie.ai pricing work?
kie.ai sells prepaid credits at $0.005 each and prices every model in credits: per 1M tokens for chat models, and per image, second or video for media models. On the kie.ai pricing API, fetched 2026-09-24, Claude Opus 5 input costs 400 credits per 1M tokens, which kie.ai converts to $2. APITariff uses that dollar figure, the usdPrice field, for every kie.ai price on this site. One chat model arrives as 2 to 4 rows (input, output, cached input, cache writes), and our engine merges input and output into one record. The same API returns an “official price” column that lists GPT-5.6 Sol at $5 and $30 per 1M tokens against OpenAI’s own $4 and $20, so APITariff does not use that column. The same credits price image models: kie.ai’s Topaz Image Upscaler bills 10 credits, $0.05, per image up to 2K and 20 credits, $0.10, at 4K, as the Topaz API pricing comparison shows, read 2026-09-24. Its video models are priced per second, and the Seedance API pricing comparison sets those rates against Replicate and fal.ai.
Is kie.ai cheaper than OpenAI and Anthropic direct?
Yes, for 31 of the 32 text models APITariff matched between kie.ai and OpenRouter on 2026-09-24. For 10 of the 11 Anthropic and OpenAI models we checked, OpenRouter’s rate equals the provider’s own standard rate. On output tokens, kie.ai’s OpenAI models cost 16% to 75% less than list, and its Anthropic models 57.2% to 71.5% less. The OpenAI outlier is GPT-5.6 Sol at 16%, because its OpenRouter rate is already $2 and $10, the rate OpenAI prints in its Batch and Flex tables. The Claude vs OpenAI API pricing comparison sets these list rates side by side.
Does kie.ai sell Claude Opus 5.5?
Not at a listed price on 2026-09-24: none of the 495 rows on the kie.ai pricing API names Claude Opus 5.5, although kie.ai had an Opus 5.5 model page on 2026-09-23. Anthropic lists Claude Opus 5.5 at $4 input and $20 output per 1M tokens. kie.ai sells the older Claude Opus 5 at $2 and $10, which is 50% of the Opus 5.5 list price and 60% under the $5 and $25 Anthropic lists for Opus 5 itself. At 1,000,000 calls of 1,000 input and 300 output tokens, Opus 5 through kie.ai bills $5,000 and Opus 5.5 direct bills $10,000, as the Claude Opus 5.5 pricing post records. The Claude Opus 5 record shows the direct and kie.ai rates on one page.
What limits and data rules does kie.ai publish?
kie.ai’s API documentation, read 2026-09-24, allows “up to 20 new generation requests per 10 seconds” per account and returns HTTP 429 above that limit. On data, it states: “Log records (text / metadata): stored for 2 months, then automatically deleted.” Generated media files are kept for 14 days, then deleted. The documentation does not say whether kie.ai trains on user data, whether credits expire, or whether failed tasks are charged, and kie.ai’s terms of use, read 2026-09-16, do not describe its relationship to the model providers. APITariff has sent no Claude, OpenAI or Gemini request through kie.ai; our one kie.ai call was a Suno request on 2026-09-16 that billed the listed 12 credits.
The rule we apply
For each model, the lower of the two per-token prices fetched on the same day wins, input and output compared separately. On that rule kie.ai is lower on both for 31 of the 32 matched models on 2026-09-24, and DeepSeek V4.1 Flash waits for a second fetch. The rule weighs price only: it does not score the 20-requests-per-10-seconds limit, the 2-month log retention or the reseller relationship, and a workload that depends on any of those should price them in.
How we matched the prices
The rule: a kie.ai model matches when its name, lowercased with every symbol removed, equals the OpenRouter model id treated the same way, so “Claude-Opus-4-8” matches anthropic/claude-opus-4.8. There is no fuzzy matching. 5 kie.ai text models have no OpenRouter match: Gemini 3 Flash, Gemini 3 Pro, Gemini 3.1 Pro (OpenAI format), GPT-5 Codex and GPT-5.4 Codex. Both feeds are fetched daily; the methodology page lists every source. For the lowest list price across all text models, read the cheapest LLM API statistics.
Drafted with AI from APITariff’s own price data, checked and edited by a person. The chart is rendered from the table above, not generated.
Change log
- First edition. 495 kie.ai price rows read from its pricing API and matched to OpenRouter list prices, 2026-09-24.
| Model | Price |
|---|---|
| Claude Opus 5Anthropic | $25.00/ 1M tokens |
| Claude Sonnet 5Anthropic | $10.00/ 1M tokens |
| Claude Haiku 4.5Anthropic | $5.00/ 1M tokens |
| GPT-6 AstraOpenAI | $50.00/ 1M tokens |
| GPT-5.6 TerraOpenAI | $12.00/ 1M tokens |
| GPT-5.6 LunaOpenAI | $1.20/ 1M tokens |
| Gemini 3.8 FlashGoogle | $3.75/ 1M tokens |
| Grok 4.7xAI | $6.00/ 1M tokens |
Affiliate link. If you sign up through it we may earn a commission. It changed nothing in the numbers above, which were fetched before the link existed.
See kie.ai pricingCite as: According to APITariff (https://apitariff.com/compare/kie-ai-vs-official-api-pricing/, data as of 2026-09-24).