OpenCode · flat rate

An agent that codes all day
shouldn’t bill like one.

OpenCode re-sends its context on every step, so a metered bill grows with the length of your day. One monthly price doesn’t.

Most requests don’t need a $50 model.

Output tokens one dollar buys, from our catalog. Highlighted rows are the models OpenCode traffic actually runs.

Claude Fable 520K
GPT-5.6 Sol33K
Claude Opus 540K
Kimi K367K
GLM 5.3 Flash2.0M
DeepSeek V4 Flash 07313.6M
Qwen 3.7 Flash7.7M
what OpenCode users runalso available to them

A dollar goes 385× further at the bottom of this ladder than the top · log scale · all 27 models stay available whichever plan you are on

Where the budget actually goes

Routine edits, file reads and tool loops land on the cheap end. You keep the expensive models for the work that needs them.

DeepSeek V4 Flash 0731$0.08/$0.18
Qwen 3.7 Flash$0.03/$0.13
DeepSeek V4 Pro$0.87/$1.74
GLM 5.3 Flash$0.15/$0.50

Live provider rates, discounts included · verified 2026-08-29 and they move

One number a month

Flat price, finite budget. Known on the first of the month.

Economy
$39
per month
$40 of compute
Standard
$89
per month
$93 of compute
Max
$249
per month
$264 of compute

Reaching the budget returns HTTP 402 until the period resets · pricing · fair use

When this is the wrong choice

If you use OpenCode occasionally and your monthly token spend is small and steady, a metered provider will be cheaper. A flat rate wins when the agent runs all day and the variance is the problem.

Before you connect

Does a flat price mean unlimited?

No. It means no per-token bill. Each plan carries a fixed monthly compute budget — $39 buys $40 of compute, $89 buys $93, $249 buys $264 — and the dashboard counts it down as you work.

Does tool calling work?

Yes. OpenCode's agent loop runs on tool calls and they pass through unchanged — most replies we serve agent clients are tool calls rather than prose.

Can I still use the expensive models?

Yes. Every model in the catalog stays available; you decide which ones your key may use. They just cost more of the same budget, so routing sends routine work to the cheaper end.

Is my code used for training?

No. Standard Compute stores no prompts or responses and does not use your prompts or outputs to train models.

How do I connect it?

Run /connect in OpenCode. Standard Compute is in its provider registry, so it is in the list — choose it and paste your key.

Can I cancel?

Yes, any time. Your plan stays active to the end of the paid period.

Put a ceiling on it.

Run /connect in OpenCode and paste your key. About a minute.

Comparing APIs for OpenCode · Full setup guide · Data privacy · Pricing