Every model we track
The whole catalogue, grouped by what the model makes. Each row says what one unit of its output costs and how many routes actually sell it.
AI video models
| Model | Creator | Cheapest | Providers | Open model page |
|---|---|---|---|---|
| t2v · i2v · ref2v released 2026-07-31 🎬 Long takes: 30 seconds in one generation, where nothing else here passes 15. | ByteDance | $2.16 derived per clip · 10s / 720p / t2v / no audio | 13 live / 13 | |
| t2v · i2v · ref2v released 2026-06-15 🔁 Volume work: drafts, variants and prompt tests, with 720p as the ceiling. | ByteDance | $0.2288 derived per clip · 10s / 720p / t2v / no audio | 11 live / 13 | |
| t2v · i2v · ref2v released 2026-02-12 📺 Resolution: the only Seedance that outputs 4K, at up to fifteen seconds. | ByteDance | $1.216 derived per clip · 10s / 720p / t2v / no audio | 12 live / 13 | |
| t2v · i2v · ref2v released 2026-02-05 🔊 Talking clips: native audio in five named languages, from 3 seconds up. | Kuaishou | $0.90 derived own unit per clip · 10s / pro / t2v / no audio | 6 live / 6 | |
| t2v · i2v · ref2v released 2025-10-15 🔗 Stitched scenes: Extend chains generations to about 148 seconds, though one clip caps at 8. | $2.56 derived own unit per clip · 8s / 1080p / t2v / audio | 9 live / 11 |
Swipe the table for price and status →
This column ranks at one shared unit: 10s / 720p / t2v / no audio. A row marked "own unit" sells nothing at it, so its figure buys what the line under it says and is not comparable with the rest of the table.
AI image models
| Model | Creator | Price | Providers | Open model page |
|---|---|---|---|---|
| t2i · i2i released 2026-06-30 also called Nano Banana 2 Lite One resolution and one price: 1K, for work that never needs more. | $0.0336 derived per image · one 1K image / t2i | 1 live / 1 | ||
| t2i released 2026-05-28 also called Nano Banana 2 Four priced resolutions from one model, so the bill scales with the size you actually need. | $0.0672 derived per image · one 1K image / t2i | 1 live / 1 | ||
| t2i · i2i released 2026-04-21 Nine published per-image prices across three sizes and three quality settings, so you pay for the picture you asked for. | OpenAI | $0.211 claimed per image · one 1024x1024 image / t2i / high | 1 live / 1 |
Swipe the table for price and status →
These models are not ranked against each other. They make the same kind of thing, and OpenAI grades it on a quality scale of its own: nothing published says which of those grades is the same picture another vendor sells, so a cheapest here would be us choosing that equivalence. Each figure below buys what the line under it says.
AI music and audio models
| Model | Creator | Price | Providers | Open model page |
|---|---|---|---|---|
| t2a released 2025-04-02 Speech billed by the script rather than by the minute, so a long read costs what it reads. | $30.00 claimed per 1M characters · 1M characters / speech | 1 live / 1 | ||
| t2a Speech billed by the length of what comes out, not by the length of what goes in. | $0.03 derived per 60 seconds of audio · 60s / speech | 1 live / 1 | ||
| t2m Music through a web app, on a plan rather than a per-song price. | Suno | — | 1 live / 1 |
Swipe the table for price and status →
These models are not ranked against each other. They do not sell a common unit that we can price them all at, so each figure below buys what the line under it says and nothing here is a cheapest.
AI text models
| Model | Creator | Cheapest | Providers | Open model page |
|---|---|---|---|---|
| text released 2026-07-24 For complex agentic coding and enterprise work, at a million tokens of context. | Anthropic | $1.00 derived Reference workload · 100K input + 20K output tokens | 1 live / 1 | |
| text released 2026-06-30 A million-token context with every rate class published, cache and batch included. | Anthropic | $0.40 derived Reference workload · 100K input + 20K output tokens | 1 live / 1 | |
| text released 2025-10-15 For high-volume work that fits inside a 200k context, with every rate class published. | Anthropic | $0.20 derived Reference workload · 100K input + 20K output tokens | 1 live / 1 | |
| text A million-token context with image input, and a stated point where the rate card changes. | OpenAI | $1.10 derived Reference workload · 100K input + 20K output tokens | 1 live / 1 |
Swipe the table for price and status →
These are prices for buying output through an API, at the unit named beside each figure. If you would rather use a model through a web interface and integrate nothing, see the consumer apps.
What ModelFare tracks
ModelFare tracks generative AI models across four kinds of output: video, images, music and audio, and text. For each one we show where you can actually buy output right now: the model creator's own API, an aggregator like fal.ai, an unofficial reseller, or a consumer app.
Every figure is quoted at one unit, and that unit is written beside it - a clip, an image, a million characters, a million tokens, or a named workload - so a price is never shown without saying what it buys. What one unit is differs by category, and two models are only ever compared at the same one. Numbers come from the provider's own published pricing and carry the date we last checked them.