The Cheapest AI APIs in 2026: Real Prices for Text, Image, Video and Music

Ask a better question than "what's the cheapest AI API?" — because nobody buys an API. You buy a specific model doing a specific job, and the cheapest text model is irrelevant if your feature generates video.

So this post lists verified per-call prices for the models people actually ship in 2026, grouped by what you're generating, then explains the four things that move your bill more than the sticker price does.

All prices USD, checked 25 July 2026 (1 credit = $0.01). "Official" is the model owner's list price where one exists.

Cheapest image generation

• Z-Image — $0.0046 per image. The cheapest call on the platform.
• Nano Banana 2 Lite — $0.0275 vs $0.0336 official (−26%)
• FLUX.2 — $0.0284 to $0.0412
• Seedream 5 Pro (1K) — $0.0412 vs $0.045 official (−17%)
• Nano Banana 2 — $0.0614 to $0.1382
• Nano Banana Pro — $0.1232 to $0.1642
• GPT Image 2 (1K) — $0.1256 vs $0.219 official (−48%)
• GPT Image 2 (4K) — $0.2493 vs $0.413 official (−45%)

At $0.0046, Z-Image is roughly 27× cheaper than GPT Image 2 at 1K. That gap is the single biggest cost lever in image work: generate drafts on a cheap model, reserve the expensive one for final output.

Cheapest video generation

Per second of output unless noted.

• Grok Imagine 1.5 image-to-video (480p) — $0.0443 vs $0.08 (−50%)
• Seedance 2.0 Mini (480p) — $0.0615 vs $0.0721 (−22%)
• Grok Imagine 1.5 image-to-video (720p) — $0.078 vs $0.14 (−49%)
• Volcengine lip sync (720p) — $0.0814
• Wan 2.7 (720p) — $0.0928 vs $0.10 (−16%)
• Kling V3 Turbo (720p) — $0.1042 vs $0.112 (−15%)
• Kling V3 Turbo (1080p) — $0.1302 vs $0.14 (−15%)
• Seedance 2.0 Mini (720p) — $0.1322 vs $0.1547 (−22%)
• Wan 2.7 (1080p) — $0.1392 vs $0.15 (−16%)
• OmniHuman 1.5 lip sync — $0.1522 vs $0.16 (−14%)
• Gemini Omni (8s, 720p) — $0.7809 per clip vs $1.00 (−29%)
• Veo 3.1 — $2.2688 to $3.3578 per clip

Video is where budgets die: one 10-second 1080p clip costs more than a thousand image generations on Z-Image. Two rules — render at 720p unless you know you need more, and keep clips short.

Cheapest text and LLM

Per million tokens, input / output.

• GPT-5.6 Luna — $0.6498 / $3.8988 vs $1 / $6 official (−41%)
• GPT-5.6 Terra — $1.6245 / $9.747 vs $2.50 / $15 (−41%)
• Claude Sonnet 5 — $1.9548 / $9.7871 vs $3 / $15 (−41%)
• GPT-5.5 — $3.249 / $19.494 vs $5 / $30 (−41%)
• GPT-5.6 Sol — $3.249 / $19.494 vs $5 / $30 (−41%)
• Claude Opus 4.7 — $3.2624 / $16.3253 vs $5 / $25 (−41%)
• Claude Opus 4.8 — $3.57 / $17.85 vs $5 / $25 (−35%)
• Claude Fable 5 — $7.14 / $35.70 vs $10 / $50 (−35%)

Note cached input if your prompts repeat: GPT-5.6 Terra cached input is $0.1624 per million versus $1.6245 uncached — a 10× reduction on the input side. Full breakdown in Claude API pricing.

Cheapest music and audio

• Gemini Omni Audio — $0.0642 per generation
• Suno V4.5 Plus — $0.1321 per generation

Suno has no official public API, which makes per-generation price the only real comparison point. See the Suno API guide.

The four things that actually decide your bill

Sticker price is where most people stop. These four move the number more.

1. Do you pay for failures? On most providers a failed, errored or empty generation is still billed. Here it's refunded automatically — which matters most on video, where one failed 10-second clip can cost more than a hundred image calls.

2. Output length and resolution. LLM output is around 5× input, and 4K is roughly 2× 720p on both image and video. Capping max_tokens and rendering at display size beat switching vendors.

3. Tier discipline. Cheap model for drafts, expensive model for the artifact the user keeps. A 3 to 5× cost difference sits in this one decision.

4. Expiring credits and minimums. Prepaid balances that expire, seat fees and monthly minimums never show up in a per-token table. Our credits don't expire and there are no seats: $5 gets you 500 credits to test with, $50 gets 5,000 plus a 20% bonus.

The cheap-stack pattern

Teams that get their AI bill under control tend to do the same four things.

• Route by difficulty. Flash and Luna-class models for the 80% that's routing, extraction and classification; premium models for the rest.
• Draft cheap, finish expensive. Z-Image or Nano Banana 2 Lite for iteration, GPT Image 2 or Nano Banana Pro for the final render.
• Measure per feature. Every create response returns creditsCharged — sum it per feature and you have unit economics without building a metering pipeline.
• Keep a fallback key. Route primary traffic to the cheaper endpoint, fall back to the official API on error.

All of it runs off one endpoint and one key, so model choice is a string:

POST https://you.bot/api/v1/generate
Authorization: Bearer YOUR_API_KEY

{ "modelId": "z-image", "input": { "prompt": "..." } }

Change modelId to gpt-image-2-text-to-image, claude-sonnet-5, kling-3-0 or suno-v4-5 and everything else stays the same. More on why in One key, every model.

Cheaper than the official APIs, per model

If you're going direct today, these comparisons are the ones worth reading next: OpenRouter alternatives, Replicate alternatives, fal.ai alternatives, Claude API pricing, Gemini API pricing, Nano Banana API and the Suno API.

FAQ

What is the cheapest AI API in 2026? It depends on modality. The cheapest single call here is Z-Image image generation at $0.0046. For video it's Grok Imagine 1.5 image-to-video at $0.0443 per second at 480p. For text it's GPT-5.6 Luna at $0.6498 per million input tokens. Prices checked 25 July 2026.

Is there a cheaper way to use GPT, Claude and Gemini than the official APIs? Yes — the same models are available through a gateway below list price. The current gap is 35 to 41% on text models and up to 50% on some media models, billed from one prepaid balance.

What's the cheapest LLM API for high volume? Look at value tiers rather than flagships: GPT-5.6 Luna at $0.6498 / $3.8988 per million tokens, and use cached input where prompts repeat.

Do cheap AI APIs mean lower quality? Not when it's the identical model. A gateway fronting the same weights returns the same output; only the price and billing terms change.

How do I avoid paying for failed generations? Choose a provider that refunds them. Here, failed, errored and empty generations are refunded automatically, so you only pay for output you receive.

Browse all models with live prices or the full catalog.