What this page answers
How much the same model costs on the official API, what aggregators charge for it, and how wide the current price gap is across purchase routes.
Video workload calculator
Estimate the batch before choosing a route
10 successful clips · 5s each · 10% retry buffer
Estimate = published price/second × generated seconds. Minimum billing, failures beyond the buffer, refunds, queue cost, and quality differences are not modeled.
Expected attempts—rounded up to whole generations
Generated seconds—duration × expected attempts
Lowest published estimate—Waiting for video pricing
Published range—same active filters, not quality-ranked
Matching listings—official and routed prices kept distinct
unit
mode
route
access
AI model purchase-channel comparison
Compare official and aggregator prices for the same AI model
CheapTokenz helps you compare purchase channels for token-priced AI models. See how much the same model costs on the official API and across aggregator routes, with a consistent view of pricing, context, and source coverage.
Quick answer: CheapTokenz is a public AI price comparison board for token-priced language models. It normalizes official API prices and aggregator route prices into the same model identity, then compares input price, output price, blended cost, context window, and source coverage. Use it as a pre-purchase routing checklist: first compare equivalent model IDs, then scan whether each row is official, aggregator, or cloud-hosted, and finally check access region, payment surface, and crawl date. The site is useful when a developer already knows the model they want to use and needs to check whether the official API, an infrastructure aggregator, or a routed provider publishes a lower comparable price. CheapTokenz does not certify providers or claim the cheapest route is automatically the best route; it keeps source freshness, provider type, and route caveats visible so the buyer can verify availability before moving production traffic.
LLM pricing
Compare public official API and aggregator route prices for token-priced language models.
Stay on LLM board
Video model pricing
Dreamina Seedance 2.0, fal Seedance, Sora, Veo, Runway, and HappyHorse benchmark costs.
Open /video/
AI coding plan comparison
Compare Codex, Claude Code, Cursor, Copilot, Gemini, Devin/Windsurf, OpenCode Go, and GLM by real coding limits.
Open /coding-harness/
Image model pricing
Compare GPT Image, Imagen, Nano Banana, FLUX, Ideogram, Grok, and image API route costs.
Open /image/
What CheapTokenz does not do
CheapTokenz compares published prices for the same model. It does not try to rank overall model quality, certify providers, or make trust claims it cannot yet justify.
How to read the table
Blended gives a practical default view of total token cost, while Input Spread isolates prompt-price differences. Together they show whether the same model is simply cheaper elsewhere.
Buyer questions before the LLM board
Start with the workload, not only the cheapest token row
The July 2 consumer-demand pass shows that LLM buyers want price translated into monthly work. The table should stay source-aware, but the decision starts with token units, output/input mix, cache and batch eligibility, context windows, rate limits, and route reliability.
Real monthly cost
Convert per-1M token prices into chatbot, summarization, RAG, agent tool-call, and batch workloads before judging value.
Output-heavy risk
Output tokens, reasoning, long answers, retries, and tool loops can move the bill faster than the input price suggests.
Route confidence
Official APIs, gateways, aggregators, and self-hosted routes should remain separate when reliability, latency, or data policy matters.
Page rule: community and SERP research informs the buyer language here; current price facts still come from the public dataset and linked provider sources.
Price gap highlights
Top price gaps today
A quick scan of a few standout like-for-like price gaps from the current public bundle. Use the language-model board for the full comparison.
GPT · 7 sources
GPT 5.5
GPT 5.5 currently resolves to 7 sources. The cheapest blended route is $2.25/M on Anyone.ai, about 98% lower than the highest priced route. It includes both official and aggregator routes.
View on language board →Claude · 10 sources
Claude Sonnet 4.6
Claude Sonnet 4.6 currently resolves to 10 sources. The cheapest blended route is $6.21/M on LingdongAPI, about 88% lower than the highest priced route. It includes both official and aggregator routes.
View on language board →Gemini · 9 sources
Gemini 2.5 Flash
Gemini 2.5 Flash currently resolves to 9 sources. The cheapest blended route is $1.3/M on AIHubMix, about 83% lower than the highest priced route. It includes both official and aggregator routes.
View on language board →DeepSeek · 8 sources
DeepSeek V4 Pro
DeepSeek V4 Pro currently resolves to 8 sources. The cheapest blended route is $3.04/M on DeepSeek, about 75% lower than the highest priced route. It includes both official and aggregator routes.
View on language board →GLM · 8 sources
GLM 4.5 Air
GLM 4.5 Air currently resolves to 8 sources. The cheapest blended route is $0.45/M on FastRouter, about 87% lower than the highest priced route. It includes both official and aggregator routes.
View on language board →Command · 3 sources
Command R
Command R currently resolves to 3 sources. The cheapest blended route is $0.487/M on OpenRouter, about 95% lower than the highest priced route. It includes both official and aggregator routes.
View on language board →
Freshness and provenance. These cards summarize normalized public prices. Check the linked provider source before buying or routing production traffic.
Bundle generated: 2026-07-14T10:09:27Z · Provider crawl window: 2026-05-25T05:49:18.000Z → 2026-07-14T10:08:27.344Z · Provider mix: 16 official, 24 aggregator or routed.
Bundle generated: 2026-07-14T10:09:27Z · Provider crawl window: 2026-05-25T05:49:18.000Z → 2026-07-14T10:08:27.344Z · Provider mix: 16 official, 24 aggregator or routed.
| Model | Blended | Input Spread | Src | Context |
|---|
Methodology
How CheapTokenz turns official and aggregator listings into a comparable table
CheapTokenz normalizes raw provider model IDs into a canonical registry so the table compares like-for-like offers instead of loosely matching vendor labels. The goal is simple: compare published pricing for the same model across different purchase channels.
Same-model matching
Rows merge priced listings only when provider IDs resolve to the same canonical model. This avoids mixing dated variants, plan wrappers, and unrelated aliases into the same comparison row.
Price reading
Blended uses input + 3×output as a practical default for total token cost, while Input Spread compares prompt pricing only. Both are route-comparison tools, not model-quality scores.
Disclosure over judgment
CheapTokenz distinguishes source types and published prices, but it does not certify suppliers or claim that a lower listed price automatically means a better route.
CheapTokenz compares price transparency, not provider trustworthiness. Some rows stay out of the main public surface when model identity, pricing scope, or route labeling are too ambiguous for an apples-to-apples comparison.
Coverage
What the homepage covers right now
The table focuses on token-priced AI model offers with enough identity confidence to compare across official APIs and aggregator routes. It is designed for quick scanning first, then row expansion when you need provider-level price cards, raw model IDs, and source links.
Families
Coverage includes major families such as GPT, Claude, Gemini, DeepSeek, Qwen, GLM, Llama, Kimi, Mistral, Command, ERNIE, Grok, and related open-weight variants when they appear with priced provider listings.
Source types
The table can include both official APIs and routing or aggregation providers. Expanding a row shows the individual price cards so you can see where the model is being sold and at what published price.
What is not implied
A lower listed price does not automatically imply better reliability, quality, or route behavior. CheapTokenz is a comparison layer for published pricing, not a certification or endorsement layer.
- Click a row to open provider cards with input price, output price, blended cost, context window, raw provider model ID, and a source link.
- Use the search box or family filters to narrow to a model family, provider cluster, or free-only subset.
- Use Blended when you want a practical total-cost comparison, and Input Spread when you care specifically about prompt price dispersion.
Dataset license: CheapTokenz publishes this normalized pricing metadata for public reference and evaluation with attribution to CheapTokenz. Provider names, model names, and source pricing pages remain subject to their respective owners and terms; verify current prices at the linked provider source before purchasing.
FAQ
Quick answers about what CheapTokenz compares and what it does not
These short explanations make the homepage easier to understand without opening every comparison row or assuming claims the site is not making.
What does Blended mean on CheapTokenz?
Blended estimates effective cost as input + 3×output and then measures the savings gap across providers for the same model.
What does Input Spread mean?
Input Spread compares only input token pricing across providers. It does not include output cost, so it is best for prompt-heavy or input-only comparisons.
Does CheapTokenz compare official APIs and aggregators together?
Yes. The table can include both official providers and routing or aggregation surfaces, but rows only merge listings that resolve to the same canonical model identity.
Are all provider listings shown publicly?
No. Some rows stay out of the main public compare surface when they are unresolved, outside the text-model scope, image-only, embedding-only, or packaged as unofficial community variants.
Does CheapTokenz certify or endorse providers?
No. CheapTokenz shows published pricing and source context, but it does not currently certify providers, assign trust scores, or make procurement guarantees.
Can I search specific models from the URL?
Yes. The homepage search box supports a shareable ?q= parameter, so a query-filtered view can be revisited or shared directly.
Video generation price board
Compare Dreamina Seedance 2.0, Sora, Veo, Runway, and video API route costs
This view keeps official APIs, cloud surfaces, and aggregator routes separate so per-second prices do not hide billing, regional, or access differences.
— model providers
— route providers
— listings
Updated —
Static /video/ page
Default sort: latest / hottest first; scenario cost breaks ties
Usable clip cost
Compare 5s and 10s costs with resolution and audio visible; one headline price rarely covers the finished clip a creator needs.
Credit burn drivers
Duration, 1080p or 4K output, pro mode, image-to-video, audio, retries, and queue priority can all change the real bill.
Production risk
Watermarks, commercial-use terms, API access, queue delays, and failure or refund rules matter as much as the per-second price.
Formula-derived and currency-normalized rows stay labeled. BytePlus Seedance uses official token formula estimates; Alibaba HappyHorse preserves CNY source pricing and FX metadata.
Loading video pricing...
Image generation price board
Compare GPT Image, Imagen, Nano Banana, FLUX, Ideogram, Grok, and image API route costs
This view keeps per-image, per-megapixel, and image-token pricing separate so cheap thumbnail routes do not get mixed with frontier image models.
— model providers
— route providers
— listings
Updated —
Static /image/ page
Units stay separate: per image · per MP · per 1M tokens
Quick answer: Image API prices should be compared by billing unit before choosing the lowest route. CheapTokenz keeps per-image, per-megapixel, image-token, edit-priced, and batch-priced rows separate because a thumbnail route, a product mockup workflow, and a frontier image model can have different cost drivers. The inline board summarizes GPT Image, Imagen, Nano Banana, FLUX, Ideogram, Grok, and related routes, while the /image/ page adds the complete table, FAQ, source notes, and billing-unit caveats for buyers who need to verify details before purchase.
Real image unit
Keep per-image, per-megapixel, token-priced, edit-priced, and batch-priced rows separate before comparing providers.
Use-case fit
Photorealism, text rendering, product mockups, style consistency, inpainting, upscaling, and bulk API workflows need different models.
Rights and privacy
Commercial use, private generations, training policy, watermarking, and marketplace disclosure risk should be visible before purchase.
OpenAI and some router rows are image-token priced; fal and Together can expose per-image or per-megapixel rows. CheapTokenz labels the unit instead of forcing a fake single-image comparison.
Loading image pricing...
loading...Blended = input + 3×output · click row ▼ compare