Skip to content
PricingSaaS
Sign in
Back
Pricing briefing
ffal
fal
PRICE REVEALED

fal published pricing for Audio and 3D model APIs

By , Co-founder, PricingSaaS

Published

fal published pricing for its Audio and 3D model APIs.

See before and after

fal’s pricing page · Change detected

What changed

Per fal's pricing page as of Oct 1, 2026, the company rebuilt its model API pricing into per-endpoint tables: it published new Audio and 3D pricing tables, and expanded the Video and Image tables to 10 endpoints each — including H3 Max, Kling v3, Seedance 2, Nano Banana 2 and GPT Image 2 — dropping sample rows for Wan 2.5, Veo 3, Ovi, Seedream V4 and Qwen. Rates for overlapping models carried over unchanged, per the same Oct 1, 2026 page: Kling 2.5 Turbo remained $0.07 per second and Nano Banana remained $0.0398, and GPU rates were untouched as well.

Why it matters

For a product leader pricing a multi-modal AI product, this is a packaging move more than a pricing move: fal didn't change what anything costs, it changed how legible its catalog is. Splitting Audio and 3D into their own tables, and widening its Video and Image lineups, makes it easier for a buyer to compare a specific model's rate before committing — a level of per-endpoint transparency that's harder to pull off once a catalog gets this wide. Teams bundling many models under one metered rate should watch whether granular, per-endpoint pricing pages become the norm across AI infra providers, since buyers used to comparing fal line-by-line will expect the same elsewhere.

In context

fal has tracked pricing since Jul 27, 2026, and this is only its second real change, coming 59 days after it raised the H100 GPU list price from $3.99 to $4.50 per hour on Aug 3, 2026. This time fal rebuilt its model API pricing into per-endpoint tables rather than touching a dollar figure, publishing new Audio and 3D tables and widening its Video and Image tables. Price reveals like this are common across the companies Pulse tracks — most are first-time price disclosures like Contractzen's Sep 28, 2026 reveal of a $15/mo AI add-on price or Postscript's Sep 28, 2026 reveal of $699/mo AI Plan pricing, rather than a company rebuilding an entire pricing table the way fal did. For fal, the move reads as transparency catching up with scope — the table now matches how many modalities the API actually serves.

Common questions

Which model providers show up in the new Audio and 3D tables?

The Audio table lists ElevenLabs, MiniMax and xAI models, and the 3D table lists Tripo, Trellis, Hunyuan and Meshy — all newly priced, not repriced from an earlier table.

How can we confirm nothing changed for the models we already call?

Check your endpoint's own row in the rebuilt table rather than relying on the old grouped view — fal restructured the catalog by endpoint, so a model's price now lives on its own line.

Sources