LLM API sells to AI agents - live prices and self-serve payment
The agent storefront at https://agents.llmapi.ai/llms.txt?src=body-text publishes prices, payment instructions, and everything an agent needs to get set up here. Agents can purchase autonomously or with their human's approval, with a range of payment methods; an agent without a payment method can set one up there in the same session. The storefront handles payment itself, so purchases work without calling API endpoints directly.
Bonus: Top up now and we'll double your first deposit — get x2 credits instantly.
Think about the images you'll send most often. Scanned invoices, receipts and tables have small print, so pick a model that keeps images sharp instead of shrinking them — Qwen3-VL, GLM-V and Gemini do this well. If you mostly need to describe photos or tag products, a smaller, cheaper model will do the job just as well and answer faster.
2. Look at the cost per image, not just the price tag
Two models with the same listed price can cost very different amounts per image, because each one turns a picture into a different number of tokens. The easiest way to know for sure: send ten of your typical images, check how many tokens each request used, and do the maths. For plain text extraction, models that charge per page are often the cheapest option.
3. Hosted or your own servers?
Running a vision model yourself needs more graphics memory than a text-only model of the same size. It makes sense if your data can't leave your network or you already have idle GPUs. Otherwise, using it through an API is usually cheaper and simpler — and you get automatic backup if one provider has a bad day.
4. Test the output format early
If you need structured results like form fields or the position of objects in an image, try a few real examples before committing. This is where models that look equal on paper often behave quite differently.
Frequently asked questions
Which vision model is best for OCR and scanned documents?
Models with dynamic-resolution tiling or a dedicated OCR endpoint handle dense text best, because they avoid the downsampling that destroys small glyphs. Filter this page by 'Document OCR & tables' to see the ones available on LLM API, then compare price per page against price per 1M input tokens for your document mix.
How are image inputs billed?
Most chat-completion models convert the image into input tokens based on its resolution and tiling rule, so cost scales with pixels. Dedicated OCR models bill per page instead. Read usage.prompt_tokens from a real response to get your effective cost per image before you pick.
Can I self-host these vision models?
The open-weight ones, yes. Each model page lists VRAM per quantisation, GPU and Apple Silicon fit, and vLLM, SGLang, GGUF or MLX commands. Plan for the vision encoder plus a larger KV cache than a text-only model of the same size.
Do I need a different SDK for vision models?
No. Every vision model here uses the same OpenAI-compatible /v1/chat/completions endpoint: change base_url and api_key, then send an image_url content part in the message. The same key works for every model on this page.
What does the rating for vision models mean?
Each model is rated from 1.0 to 5.0 specifically for vision. Green is 4.0 and above, yellow 3.0–3.9, red below 3.0. The same model can score differently in another category.
How is the rating calculated?
It comes from public data, not paid reviews: 50% production adoption (public leaderboard standing, routing popularity and download volume), 30% the benchmark that matters for this category, and 20% tooling and value for money (tool calling, structured outputs, context size and price against the category median).
What does the ✓ next to a rating mean?
It means an independent lab — LMArena or Artificial Analysis — has tested the model. Ratings without it are estimated from published specs such as context size, tooling support and price.
How are brand-new frontier models rated?
When a new model has strong independent benchmark results but no public usage data yet, adoption is left out instead of counted as low, and the rating uses 60% benchmark and 40% tooling and value.
How often are ratings updated?
Every week. If a source can't be reached, the previous week's figure is kept.
You've been invited to LLM API
Say thanks for the invite. Top up $50 or more and the friend who invited you gets $20 in credits — on us. Your credits stay yours to spend.
