Pricing
Pay for the
tokens you use.
Every request is metered to the token. We charge the provider's per-million rate plus 25% — and we show you both numbers, every time. No subscriptions. No seat fees. No minimum.
How billing works
Top up
Buy a credit pack — each credit is $1 of usage
Make requests
Each call charged: tokens × provider rate × 1.25
See the cost
Real provider cost + our margin, on every response
Credit packs
Same per-token rate regardless of pack size. The packs are just top-up amounts — your real cost is per-call. Packs are priced in dollars; checkout shows the equivalent in your currency. Credits stay valid for 18 months from purchase.
Per model
What does $25 get you?
All prices below include our 25% margin. The "$25 buys" column shows input tokens you can send for $25 at that model's rate.
| Model | In $/M | Out $/M | $25 buys (input) | Best for |
|---|---|---|---|---|
Claude Fable 5.1 flagship 1M context | $12.50 | $62.50 | 2M input tokens | The frontier. Use it when it has to be right. |
Claude Opus 5.5 new 1M context | $5.00 | $25.00 | 5M input tokens | Frontier quality at under half of Fable's rate. |
Claude Sonnet 5.5 new 1M context | $2.50 | $12.50 | 10M input tokens | The workhorse. Best balance of speed + quality. |
Claude Haiku 5.5 value 1M context | $0.125 | $0.625 | 200M input tokens | Reviews, formatting, sub-agents. Prompts over 100K tokens bill 5×. |
GPT-6 Astra 1M context | $12.50 | $62.50 | 2M input tokens | OpenAI's flagship. Try it next to Fable 5.1. |
GPT-6 Sol 1M context | $2.50 | $12.50 | 10M input tokens | GPT-6 for complex coding and agentic work. |
GPT-6 Luna value 1M context | $0.125 | $0.625 | 200M input tokens | The efficient GPT-6 — focused, high-volume tasks. |
Grok 4.7 new 500K context | $2.50 | $7.50 | 10M input tokens | xAI's flagship — agentic coding and real-world work. |
Grok 4.1 Fast 2M context | $0.25 | $0.625 | 100M input tokens | Massive-context (2M) work at budget rates. |
Gemini 4 Argon new 1M context | $5.00 | $25.00 | 5M input tokens | Google's newest frontier model (promotional rate to 30 Nov). |
Gemini 3.8 Flash 1M context | $1.88 | $9.38 | 13M input tokens | Google's most capable Flash — long-horizon work. |
Gemini 3.5 Flash-Lite 1M context | $0.375 | $3.13 | 67M input tokens | High-volume pipelines at minimal cost. |
GLM-5.3 1M context | $1.75 | $5.50 | 14M input tokens | Z.ai's flagship. Always thinks. |
Qwen 3.8 Max 1M context | $2.50 | $7.50 | 10M input tokens | Alibaba's flagship MoE. |
DeepSeek V4 Pro value 1M context | $1.65 | $4.95 | 15M input tokens | Top open-weight reasoning at a fraction of frontier rates. |
DeepSeek Flash value 1M context | $0.375 | $1.50 | 67M input tokens | Budget reasoning for throwaway calls. |
Cost = (input tokens × in $/M) + (output tokens × out $/M). Output tokens are typically 2–6× more expensive than input on most providers.
Plus image, video, voice, and TTS models priced per asset. Full table in the docs.
Compute
Run code, not just prompts
The assistant can build and test in an isolated cloud sandbox — full toolchain, destroyed after each run — or on your own machine, where your code never leaves your computer. Compute is metered by the second only while a command runs.
Your machine
Connect the bridge and the assistant builds on your own hardware — your data never leaves your computer. You only pay for the models it calls.
Cloud sandbox
An isolated container (2 vCPU · 4 GB), prompt-injection-safe. Stays warm while you work so incremental builds are fast; reaped after an idle hour. Billed per second of run time — a typical build + test is a fraction of a cent.
Coding session
A full autonomous coding agent in a bigger container (up to 8 vCPU · 32 GB) that works a whole task end-to-end. Priced by container tier + model usage.
All compute runs on Quantum Encoding's own multi-region infrastructure. Credits cover models and compute alike — one balance, one bill.
Ready to build?
Sign up, top up from $5, and see the real per-call cost on your first response.
Get startedFAQ
Frequently asked questions
How am I actually charged?
Per request, per token. Every API call is metered: input tokens × the per-million input rate, output tokens × the per-million output rate. We charge the provider rate plus our margin — 25% on the Standard tier, on every provider. The exact cost shows in the response headers and on every entry in your dashboard. No rounding up, no per-task fudge factors.
Why credits then?
Credits are just a top-up balance — $1 = 1 credit. Your balance ticks down as actual API costs accrue. Nothing more clever than that. We don't pretend a 'credit' equals a fixed amount of compute, because it doesn't — different models cost wildly different amounts per token.
Do credits expire?
They stay valid for 18 months from the date you buy them. In practice that is far longer than anyone takes to spend a top-up, and it means we are not carrying a balance on the books indefinitely.
Can I choose which model runs each call?
Yes. Pick your model per call, per mission, per task. Mix providers — Fable 5.1 for planning, Sonnet 5.5 for execution, Haiku 5.5 or DeepSeek Flash for cheap throwaway calls. Your call. We just route the request and bill you the real cost plus margin.
What happens if I run out of credits mid-mission?
The mission pauses. Top up and resume exactly where you left off. No work lost, no restarts.
Why the 25% margin?
Because the orchestration engine, multi-provider routing, sessions, billing, observability, and the live workspace aren't free to build or run. Every API response shows the raw provider cost and our margin separately — no obfuscation. If you're spending more than $50 a month, apply for the Developer tier and the margin halves to 12.5%. See the developers page.
Is there a free tier?
No. Accounts start at zero. Packs start at $5, and every call shows exactly what it cost, so a first experiment costs pennies.