AI API Pricing Comparison
Find the cheapest AI API for the model you already use. Same job, priced up to 3.3× apart across platforms.
1K
Low $0.012 · high $0.040
3.3×
Price gap
- APIMart$0.012
- Replicate$0.034
- WaveSpeed$0.036
- fal$0.040
Green: cheapest exact match · Striped: approximate
Green: cheapest exact match · Striped: approximate
Full comparison- Platforms
- 6
- Endpoints listed
- 4,165
- Current prices
- 8,456
- Model families
- 47
Compare AI API prices by task
Every endpoint grouped by task, whoever made the model, with the cheapest exact price today and who charges it.
- Text to video465 endpointsCheapest$0.005/s
- Image to video423 endpointsCheapest$0.005/s
- Video to video & editing172 endpointsCheapest$0.005/s
- Lip sync & talking avatars85 endpointsCheapest$0.005/s
- Video upscaling & enhancement44 endpointsCheapest$0.0013/s
Today’s biggest AI API price gaps
The same model and the same job, priced on every platform that sells it. Green is the cheapest exact match.
- Nano Banana 2.11K · on 4 platformsAPIMart$0.012$0.040fal3.3×Save 70%
- Gemini 3.6 FlashChat · 3:1 in/out · on 3 platformsOpenRouter$0.791$2.40APIMart3.0×Save 67%
- Claude Opus 5Chat · 3:1 in/out · on 4 platformskie.ai$4.00$10.55OpenRouter2.6×Save 62%
- Veo 3.1 Fast8 s · 720p · With audio · on 5 platformsAPIMart$0.64$1.20WaveSpeed1.9×Save 47%
- Seedance 2.05 s · 720p · on 3 platformsReplicate$0.90$1.51fal1.7×Save 40%
- Seedream 5Pro · on 3 platformsReplicate$0.090$0.135fal1.5×Save 33%
Show the full comparison table (11 models)
| Model · job | kie.ai | APIMart | fal | Replicate | WaveSpeed | OpenRouter | Spread |
|---|---|---|---|---|---|---|---|
| Veo 3.1 Fast 8 s · 720p · With audio | $0.30 | $0.64Official≈ $0.14 · Ext | $1.20 | $1.20 | $1.20 | $0.844Vertex | 1.9× |
| Kling 4.0 5 s · 720p | $0.713 | $0.70 | $0.785 | $0.785 | $0.667 | — | 1.2× |
| Seedance 2.0 5 s · 720p | “From” only | “From” only | $1.51 | $0.90 | $1.08 | — | 1.7× |
- $0.64
- Cheapest exact match
- ≈ $0.14
- Approximate: does the job with a difference
- —
- Not listed
Six AI API platforms at a glance
How much each one sells, how often it has the cheapest price in a model family, and when it was last read.
- OpenRouter551 endpoints · Cheapest in 13 of 44 families13 / 44
- APIMart358 endpoints · Cheapest in 11 of 44 families11 / 44
- WaveSpeed1,133 endpoints · Cheapest in 7 of 44 families7 / 44
- Replicate434 endpoints · Cheapest in 6 of 44 families6 / 44
- fal1,464 endpoints · Cheapest in 5 of 44 families5 / 44
- kie.ai225 endpoints · Cheapest in 2 of 44 families2 / 44
8,456 prices on 4,165 listings · last collected 15 h ago · Collector status
Latest price changes
Cuts and rises on the same listing, newest first. New and delisted models are on the changes page.
- Price cut33%Z.ai: GLM 5.2OpenRouter · Cached input (read)OpenRouter · Cached input (read)$0.009 → $0.006/M cache read
- Price cut30%Z.ai: GLM 5.2OpenRouter · Input tokensOpenRouter · Input tokens$0.01 → $0.007/M in
- Price cut38%Tencent: Hy3OpenRouter · Cached input (read)OpenRouter · Cached input (read)$0.033 → $0.020625/M cache read
- Price cut38%Tencent: Hy3OpenRouter · Output tokensOpenRouter · Output tokens$0.528 → $0.33/M out
- Price cut38%Tencent: Hy3OpenRouter · Input tokensOpenRouter · Input tokens$0.132 → $0.0825/M in
- Price cut84%Z.ai: GLM 5.3 FlashOpenRouter · Output tokensOpenRouter · Output tokens$0.5 → $0.0784/M out
- Price cut73%Z.ai: GLM 5.3 FlashOpenRouter · Input tokensOpenRouter · Input tokens$0.15 → $0.0399/M in
How it works
No prices typed in by hand, and none overwritten.
- 01
Read every public price
A collector reads the pricing APIs and pages of six platforms, daily by default, and keeps every response byte for byte.
- 02
Put them in one unit
Credits become US dollars at the platform’s own rate. Each price is tagged by task, resolution, duration and audio, so like is compared with like.
- 03
Compare, and keep the history
The cheapest exact price wins; approximate matches are marked and never ranked. When a price changes, the old one is closed, not erased.
- No referral links, no paid placement
- Every price links to the response it came from
- Full price history for every model
- Read-only API and MCP, free
For developers and agents
Every page reads as Markdown, every price is in a JSON API, and an MCP server lets Claude or Cursor ask the ledger directly.
- “Copy for agent” on every model, family and task page
- Add .md to any URL, or send Accept: text/markdown
- JSON API: search, cheapest, one model — no key
- /llms.txt indexes the whole site for language models
# MCP (Claude Code)claude mcp add --transport http apipricing https://apipricing.codertony.com/mcp # Markdowncurl https://apipricing.codertony.com/video/image-to-video.md # JSONcurl "https://apipricing.codertony.com/api/v1/cheapest?task=video/image-to-video"
AI API pricing FAQ
The full rules for every number are written down. Read the methodology →
Which platform is cheapest for AI video, image or LLM APIs?
It depends on the model and the job; no platform is cheapest for everything. Each model family and task page shows the cheapest exact price on every platform, and the home page lists today’s widest gaps for the same model and job.
Where do the prices come from?
From each platform’s own public pricing API or pricing page, read automatically, daily by default. Every response is archived, and each model page links to the response its prices came from.
How are credits converted to US dollars?
At the rate the platform publishes for a new, ordinary account, for example 1 kie.ai credit ≈ $0.005. Top-up bonuses and membership tiers are not applied to list prices.
What do “exact” and “approximate” mean?
Exact means the endpoint does exactly the job priced, such as an 8-second 720p clip with audio. Approximate means it can do the job with a stated difference, such as only making 5-second clips. Only exact prices are ranked and shown in green.
Does this site earn from the platforms it lists?
No. There are no referral links, partnerships or paid placements. Platform links go to the platform’s own pages.
Can my agent or code read these prices?
Yes, for free and without a key: add .md to any page, use the JSON API at /api/v1, or connect the MCP server at /mcp. The agent guide has examples.
Find the cheapest way to call a model.
4,165 endpoints and 8,456 prices, read from six platforms.