Question 1

Is Pier open source?

Accepted Answer

Yes. Pier Code is open source and developed in the open on GitHub at github.com/alphabench/pier. You can read the code, build it yourself, file issues, and send pull requests. Bring your own Sarvam or OpenRouter key and run it free, or subscribe to Pier for every model under one login.

Question 2

Does Pier run locally?

Accepted Answer

Yes. The CLI runs as a single binary on your machine. In local-only mode your code never leaves your environment; cloud model calls are opt-in and clearly indicated.

Question 3

Which languages and frameworks are supported?

Accepted Answer

Any text-based codebase: TypeScript, Python, Go, Rust, Java, and more. It learns your project's conventions rather than assuming a stack.

Question 4

What's included in the $5 Pro plan?

Accepted Answer

Pro is $5 a month and gives you the full Pier agent with every model included: the Pier Hybrid router, Sarvam's models, DeepSeek, GLM, Kimi, Qwen and more, each billed at its API cost. It comes with a monthly pool of usage credits that covers typical day-to-day coding.

Question 5

What happens if I run out of credits?

Accepted Answer

Buy a $5 usage pack any time, even mid-session. Packs stack, charge the same per-token rates as Pro, and the credits don't expire, so you only pay for what you actually use beyond the monthly pool.

Question 6

How can Pier be this much cheaper?

Accepted Answer

Models are billed at their providers' API list prices, with cost-effective options like Sarvam's models and DeepSeek priced far below frontier US models. The Pier Hybrid router keeps everyday work on the low-cost models, and a coding agent reads far more than it writes, so those input savings compound across every repo read, diff, and tool result.

Question 7

Is a cheaper model good enough for real coding?

Accepted Answer

For everyday work (planning, codebase Q&A, refactors, test fixing), the cost-effective models Pier Hybrid routes to hold their own and are competitive on standalone coding benchmarks like LiveCodeBench. When a task genuinely needs more, switch to a higher-capability model like GLM 5.2 or Kimi K2.7 in the picker; it is billed at that provider's API cost.

Question 8

Which models can I use?

Accepted Answer

Every model in the picker is available on every plan, each billed at its provider's API cost from your credits: the Pier Hybrid router, Sarvam 105B and 30B, DeepSeek V4 Pro and Flash, GLM 5.2, Kimi K2.7 Code, Qwen3 Coder and Coder Flash, Codestral, and MiniMax M3. Reach for a higher-capability model on the hardest tasks and fall back to the Pier Hybrid router for everyday work.

Question 9

Can I bring my own API key?

Accepted Answer

Yes. Pier is free to use with your own key: bring a Sarvam AI or OpenRouter key and Pier calls that provider directly, so you pay the provider, not us. All it needs is a free Pier account. With an OpenRouter key you can reach any model OpenRouter exposes. If you'd rather not manage keys, a Pier Pro or Max subscription bundles every model under one login.

Question 10

What does Pier send off my machine?

Accepted Answer

Only the model prompt for the calls a task actually needs: the slices of code, diffs, and tool output the agent reads to reason about your task, sent to the model provider over TLS. In local-only mode nothing leaves your machine. Cloud calls are opt-in and clearly indicated before they run.

Question 11

Where does my data go when I pick a third-party model?

Accepted Answer

Alphabench does not build models. We are a gateway that routes your prompts to the model platform you select (Sarvam AI, DeepSeek, OpenRouter, or Fireworks AI). The data we store is never used to train any model by Alphabench. However, each provider has the right to process data under its own data and compliance policies, so pick the provider whose policies fit your project.

Question 12

How is model usage billed?

Accepted Answer

Every model is billed at its provider's API cost, deducted from your plan's usage credits. There is no per-model markup on top of the API price, and the model picker shows this on each model.

Questions, answered

More on this

Pricing

Cost per task

Security

Begin here