Clear before you call
Every model on TokLab lists its live price, its parameters and the request format it accepts, so you know the cost and shape of a call before you write it.
TokLab – one API for every AI model, billed only when it works
Call AI image, AI video, LLM and AI audio models with one TokLab key. Check the price before you send, get the same error codes from every model, and pay nothing for requests that fail.
- No subscription
- Failed requests cost $0
- Credits never expire
- OpenAI-compatible
Model providers on TokLab
- OpenAI
- xAI
Google
- Anthropic
ByteDance
MiniMax
DeepSeek
Kimi
Meta
Kuaishou
/ 01—Why TokLab
Most AI APIs bill the attempt. TokLab bills the result.
A timeout, a filtered prompt or an empty render still costs money on most platforms. On TokLab a failed task is free, so retries, fallbacks and experiments never cost you twice.
One integration, every model
TokLab speaks the OpenAI, Anthropic and Gemini formats, plus one task API for media. Moving to another TokLab model means changing one field, not adding an SDK.
Credits never expire
Your balance has no expiry date and no monthly reset. Top up once and spend it this week or next year. There is no subscription to cancel.
/ 02—Models
Explore models by what they make
Video, image, language and audio models on TokLab share one key, one balance and one way to read results.
Video16 models
AI video API
Turn a prompt or a still image into video, many models with sound in the clip. Seedance 2.5, Veo 3.1, Kling 4.0 and MiniMax H3 all run on the TokLab task API.
- Text to video
- Image to video
- Native audio
Image9 models
AI image API
Generate images from a prompt, or edit one with reference images and plain instructions. GPT Image 2.5, Nano Banana, Midjourney and FLUX return files stored on TokLab.
- Text to image
- Image editing
- Reference images
LLM18 models
LLM API
Chat, reasoning, coding and agents with GPT-6, Claude, Gemini, Kimi and DeepSeek. Keep your SDK: TokLab accepts Chat Completions, Responses, Messages and Gemini requests.
- 4 request formats
- Streaming
- Price tiers
Audio1 model
AI audio API
Transcribe speech, or translate it into English, with Whisper-1 on TokLab. Billed by the minute and returned as JSON, SRT or VTT.
- Transcription
- Translation
- SRT and VTT
/ 03—Integration
From sign-up to first call in 4 steps
Keep the SDK you use today. Change the base URL and the key, and TokLab handles billing, tasks and storage.
- 01
Create an account
Sign in to TokLab with Google. Every model and its price are visible before you add any money.
- 02
Add credits
Top up in USD by card, PayPal or crypto. Credits work on every TokLab model and never expire.
- 03
Create an API key
Make one key per app, and revoke any of them from the TokLab dashboard.
- 04
Send a request
Point your client at the TokLab base URL and pick a model. The Quickstart has a working example for each request type.
import osfrom openai import OpenAI client = OpenAI( base_url="https://api.toklab.ai/v1", api_key=os.environ["TOKLAB_API_KEY"],) reply = client.chat.completions.create( model="claude-opus-5-5", messages=[{"role": "user", "content": "Hello from TokLab"}],)print(reply.choices[0].message.content)
/ 04—Features
The parts you would otherwise build yourself
Billing, media tasks, errors and usage behave the same on every TokLab model, so the code around your model calls stays small.
01
Retries that cost nothing
A failed or rejected task costs nothing, so a retry never bills you twice.
> task failed → $0> output delivered → billed once> credits → never expire
02
One task API for media
Image, video and audio jobs return a task id at once. Poll it, then download the file from TokLab storage.
POST /v1/videos/generations → idGET /v1/tasks/{id} ← statusdone ← output URL
03
Errors you can branch on
The same failure returns the same code on every TokLab model, each with a clear next step.
30001 → add credits30002 → back off, then retry60001 → revise the prompt
04
Price tiers on one key
Some models come in more than one price tier. Name the tier in the model field; TokLab never switches it behind your back.
model: gpt-image-2model: gpt-image-2:<tier>no silent tier switch
05
Agents included
Run Claude Code or Codex on TokLab models, or connect any MCP client to list models, read prices and submit jobs.
ANTHROPIC_BASE_URL=https://api.toklab.aicodex -m gpt-6-astramcp → api.toklab.ai/v1/mcp
06
Usage you can audit
Every request lands in your TokLab usage history with its charge, next to task logs and LLM logs.
usage → charge per requesttasks → status and outputllm logs → tokens in and out
/ 05—Pricing
Every price, before you call
Live rates from TokLab billing, the same numbers you are charged. Filter by what the model makes.
GPT Image 2.5 is OpenAI's image model, released as Flare and Sunburst: text to image and edits with up to 16 references, six quality levels, up to 3840 px.
Text to ImageImage to Image$0.00448 per image
Muse Image is Meta's image model. Create pictures from text or up to 9 reference images in seven aspect ratios. One price per image, no subscription.
Text to ImageImage to Image$0.0048 per image
Muse Video is Meta's video model. Turn a prompt or up to 9 reference images into a 10-second 720p clip with sound. One price per clip, no subscription.
Text to VideoImage to Video$0.024 / clip
Try GLM-5.3 for coding, long document review and tool workflows. Explore reasoning controls, live token pricing and complete developer examples.
Text Generation$1.12/1M input$3.52/1M output$0.208/1M cache read$1.12/1M cache write
Try Kimi K3.1 Pro for coding, document review and tool workflows. Explore its million-token context, reasoning controls and live token prices.
Text Generation$1.60/1M input$8.00/1M output$0.16/1M cache read$1.60/1M cache writeMore tiers →
PixVerse V6 makes 1–15 second clips from text, a photo, two frames or reference images, up to 1080p with optional audio. Pay per second; failed runs are free.
Text to VideoImage to Video$0.02 per second
PixVerse V6 Max turns a prompt or one photo into a 1–15 second clip at up to 1080p, with optional audio. Pay per second, no plan; failed runs are free.
Text to VideoImage to Video$0.024 per second
Grok Imagine Video 1.5 turns a prompt or one image into a 6–15 second clip with sound at 480p or 720p. Pay per second, no plan, failed runs are free.
Text to VideoImage to Video$0.04 per second
Kling 4.0 brings 30-second video, 10 keyframes and stereo sound. Explore its features, pricing and Flash differences. Now available on TokLab.
Text to VideoImage to Video$0.1425 per second
Create and edit images with FLUX 3 Image Klein. Combine up to ten references, choose output up to 4K, and check live prices before generating on TokLab.
Text to ImageImage to Image$0.00753 per image
MiniMax H3 makes 4–15 second clips at 768P or 2K with audio from text, frames or reference images, video and audio. Official request format, pay per second.
Text to VideoImage to Video$0.01539 per second
Create MiniMax H3 Max videos from text, frames or reference media. Choose 480P or 768P, control prompt expansion, and see live prices before you generate.
Text to VideoImage to Video$0.0457 per second
Create short videos with MiniMax H3 Max Turbo from text or first and last frames. Try 480P or 768P, compare live prices, and see a real generated sample.
Text to VideoImage to Video$0.02285 per second
Try MiniMax M3.1 Flash with text, images and video. Compare live token prices, control reasoning effort, and connect your app with three familiar API formats.
Text Generation$0.09/1M input$0.36/1M output$0.018/1M cache read$0.09/1M cache writeMore tiers →
Transcribe recordings with Whisper-1. Get readable text, word timestamps and subtitle files, check live pricing, and try your own audio online on TokLab.
Speech to text$0.036 / minute
Explore Midjourney V8.2 for visual concepts, style references and HD images. Compare versions, check task prices and follow a documented image workflow.
Text to ImageImage to Image$0.08 / job
Use Gemini 3.1 Pro for coding and multimodal analysis. Explore thinking levels, a 1M-token input window, native requests and live token prices.
Text Generation$0.90/1M input$5.40/1M output$0.09/1M cache readMore tiers →
Use Gemini 3.8 Flash for multimodal reasoning and structured text. Explore native requests, thinking levels, live group prices and the Playground.
Text Generation$0.675/1M input$3.375/1M output$0.0675/1M cache readMore tiers →
Gemini 4 Argon is Google's frontier model for long-horizon coding, legal and finance work, and cyber defense, with 1M-token output. Release status, benchmarks.
Text Generation$0.90/1M input$4.50/1M output$0.045/1M cache readMore tiers →
Create and edit images with GPT Image 2 on TokLab. Use text prompts, reference images and transparent backgrounds, with a live playground and usage-based pricing.
Text to ImageImage to Image$0.00425 per image
Nano Banana 2.1, Google's update to Nano Banana 2: create and edit images with sharper text, 1K to 4K, 14 aspect ratios and 14 references, one price per image.
Text to ImageImage to Image$0.021 per image
Try Grok 4.7 for coding, image analysis and research. Compare live token prices, choose reasoning effort and test Chat, Responses or Messages with one key.
Text Generation$1.20/1M input$3.60/1M output$0.30/1M cache readMore tiers →
Try Claude Haiku 5.5 for classification, extraction and summaries. See live token prices, test your prompts, and connect your app with the Messages API.
Text Generation$0.035/1M input$0.175/1M output$0.0035/1M cache read$0.04375/1M cache writeMore tiers →
Try Claude Sonnet 5.5 for code reviews and document questions. See live token prices, test a prompt, and check verified capabilities before connecting your app.
Text Generation$0.70/1M input$3.50/1M output$0.035/1M cache read$0.875/1M cache writeMore tiers →
Call Claude Fable 5.1 with the Anthropic or OpenAI SDK. Deep reasoning, 1M-token context, cheap cached reads, live pricing and a playground to try it today.
Text Generation$3.50/1M input$17.50/1M output$0.0875/1M cache read$4.375/1M cache writeMore tiers →
Call Claude Fable 5 with the Anthropic or OpenAI SDK. Deep reasoning, 1M-token context, forced tool use, live per-token pricing and a playground to test it.
Text Generation$3.50/1M input$17.50/1M output$0.35/1M cache read$4.375/1M cache writeMore tiers →
Call Claude Opus 5.5 with the Anthropic or OpenAI SDK. 1M-token context, adaptive thinking, live per-token pricing and a playground you can try right now.
Text Generation$1.40/1M input$7.00/1M output$0.07/1M cache read$1.75/1M cache writeMore tiers →
Call GPT-6 Astra with the OpenAI SDK. OpenAI's most capable model, with a 1.05M-token context, live per-token pricing and a playground you can try right now.
Text Generation$1.00/1M input$5.00/1M output$0.10/1M cache read$1.25/1M cache writeMore tiers →
Call GPT-6 Sol with the OpenAI SDK. 1.05M-token context, reasoning effort from none to max, live per-token pricing and a playground you can try right now.
Text Generation$0.20/1M input$1.00/1M output$0.02/1M cache read$0.25/1M cache writeMore tiers →
Call GPT-6.1 Sol with the OpenAI SDK: near-Astra coding and agent work at a fifth of the price, 1.05M-token context, 95% off cached input. Try it now.
Text Generation$0.20/1M input$1.00/1M output$0.01/1M cache read$0.25/1M cache writeMore tiers →
Call GPT-6 Luna with the OpenAI SDK. The lowest-cost GPT-6 model, with a 1.05M-token context, live per-token pricing and a playground you can try right now.
Text Generation$0.01/1M input$0.05/1M output$0.001/1M cache read$0.0125/1M cache writeMore tiers →
Call Kimi K3 with the OpenAI or Anthropic SDK. 1M-token context, always-on reasoning with low, high or max effort, live per-token pricing and a playground.
Text Generation$2.40/1M input$12.00/1M output$0.24/1M cache read$2.40/1M cache writeMore tiers →
Call DeepSeek V4.1 Flash with the OpenAI or Anthropic SDK. 1M-token context, thinking on or off, image input, live peak and off-peak pricing and a playground.
Text Generation$0.24/1M input$0.96/1M output$0.0048/1M cache readMore tiers →
Kling 3.0 makes 3–15 second clips from text or a first and last frame, with optional native sound, up to five shots and up to 4K. Pay per second, no plan.
Text to VideoImage to Video$0.08064 per second
Kling 3.0 Turbo makes 3–15 second clips with sound from a prompt or a first frame, at 720p or 1080p. Pay per second, no plan, failed runs free.
Text to VideoImage to Video$0.13728 per second
Veo 3.1 makes 4–8 second clips at 720p, 1080p or 4k from a prompt or images. Pay per clip, or per second with audio, seed and more. Failed runs are free.
Text to VideoImage to Video$0.0105 per second
SD Video turns a prompt, a first frame or your own references into a 5–15 second clip with matching dialogue and sound. 480p or 768p, six aspect ratios.
Text to VideoImage to Video$0.02 per second
Generate video with Seedance 2.0: text, first and last frames, and image, video or audio references. 480p to 4K, 4–15 seconds. Pay per second, no subscription.
Text to VideoImage to Video$0.07027 per second
Generate, edit and extend video with Seedance 2.5: up to 30 seconds at 1080p, with up to 30 image, 10 video and 10 audio references. Pay per second of video.
Text to VideoImage to Video$0.10262 per second
Generate video with Seedance 2.0 Fast: the same inputs as Seedance 2.0 at a lower price per second, at 480p or 720p for 4–15 seconds. Pay per second of video.
Text to VideoImage to Video$0.05613 per second
Generate video with Seedance 2.0 Mini, the lowest price per second in the Seedance 2.0 family: text, frames and references at 480p or 720p, 4–15 seconds.
Text to VideoImage to Video$0.03745 per second
Create and edit images with Nano Banana 2 (Gemini 3.1 Flash Image): 512px to 4K, 14 aspect ratios, up to 14 references. Pay per image or per token.
Text to ImageImage to Image$0.01875 per image
Create and edit images with Nano Banana Pro (Gemini 3 Pro Image): 1K to 4K output, built-in thinking, up to 14 references. Pay per image or per token.
Text to ImageImage to Image$0.01875 per image
Create and edit images with Nano Banana 2 Lite (Gemini 3.1 Flash Lite Image), Google's fastest and cheapest image model: 1K output, 14 aspect ratios.
Text to ImageImage to Image$0.02565 per image
Prices are charged per request against your balance. A request that fails is not charged. A model priced in tiers is listed at its standard rate, and a longer request can settle at a higher tier.
See all 44 models/ 06—FAQ
Common questions
What developers ask before they build on TokLab.
01What is TokLab?
TokLab is an AI model API platform. One TokLab key calls AI image, AI video, LLM and AI audio models, 60+ in all, through one base URL. You top up a USD balance, see each model's price before you call it, and pay only for requests that succeed.
02How does TokLab pricing work?
Pay as you go, with no subscription and no monthly minimum. Each model is billed in its own unit: per million tokens, per image, per second of video or per minute of audio. Rates on TokLab come live from billing, a failed task costs $0, and credits never expire.
03How long does it take to switch to TokLab?
If your code already calls an OpenAI, Anthropic or Gemini endpoint, change the base URL and the key and keep everything else. Image, video and audio models share one TokLab task API: submit the request, get a task id, then poll for the result.
04Which AI models can I use on TokLab?
Video models such as Seedance 2.5, Veo 3.1 and Kling 4.0; image models such as GPT Image 2.5, Nano Banana and Midjourney; language models such as GPT-6, Claude Opus 5.5 and Gemini; and Whisper-1 for speech. The TokLab models page lists every one with its live price.
05What happens when a request fails?
TokLab does not charge for it. The response carries an error code that means the same thing on every model, with a next step: retry, fix a parameter, add credits or try another model. Every model uses the same key, so falling back to another one is a one-field change.
06Can I call the same model for less?
Some models on TokLab come in more than one price tier. Add the tier to the model name, for example gpt-image-2:<tier>, and the same key bills the lower rate. A tier request never moves to another tier, so you are never billed at a price you did not pick.
07Can AI agents use TokLab?
Yes. Claude Code and Codex CLI can run on TokLab models, billed to your balance. The TokLab MCP server also lets Claude Code, Cursor and other MCP clients list image and video models, read their prices and submit jobs with your key.
08How do I pay for TokLab?
Top up by card, PayPal or crypto; crypto top-ups add a 2% bonus. Turn on Auto Top-Up and TokLab charges your saved card when the balance drops below a threshold you choose, so a running app never stops for lack of credits.
/ 07—Blog
Latest from the blog
Guides, comparisons, and deep dives on AI image, video, and model APIs.
AI photo editing
Best AI image editor in 2026: 4 models, 5 everyday edits, measured
Four AI image editors, five prompt edits. All of them changed the price on a chalkboard correctly. The cheapest left the rest of the photo almost untouched.
AI image models
The best AI image generators in 2026: 8 models, the same 4 prompts
Most AI image apps run one of eight models. We gave all eight the same four prompts. Six wrote a chalkboard sign exactly right; the cheapest cost half a cent.
AI video models
The 7 best AI video generators in 2026, tested on one rainy street
We sent one street-musician prompt to seven AI video models. One cut to close-ups, one invented a street sign, and the cheapest clip cost under eight cents.
Image to video
Best image to video AI in 2026: one photo through 5 models
One photo of a woman in the rain, five image to video models. All five kept her face. One lowered the umbrella as asked, and one put the photo in black bars.
Get started
Ship your next AI feature on TokLab
One key for AI image, AI video, LLM and AI audio models. Top up in USD and pay only for requests that succeed.
- No subscription
- Credits never expire
- Failed requests cost $0