Open at the core.
Open-weight models are the foundation. Our goal is to make model choice part of your workflow, not a decision made for you.
The model philosophyMeet Ablitron Code: our planned terminal coding experience with abliterated models. One subscription funds your coding workflow and direct API requests.
Both options fund the same API balance.
Proposed plans from $20/month include $22 in API credit. Use either model; the lower rate stretches your balance further.
Explore proposed plans →ON DEMAND / PLANNEDPrepaid API usage without a subscription. Separate input and output rates make consumption visible.
Explore the API →In development. Neither subscriptions nor API access are available to purchase yet.
Selected for launch.
Quality benchmarks come next.
The lower-cost option. Base checkpoint undisclosed; coding quality and speed are still being evaluated.
Proposed: $2 input / $6 output per 1M tokens ↗PRIVATE PREVIEWBased on a modified GLM-5.3. Our GLM-5.3-based option for coding and reasoning. Tool calls and streaming passed initial live checks; coding benchmarks are pending.
Proposed: $6 input / $10 output per 1M tokens ↗Both models. One balance.
More credit with every subscription.
$20/month
$22 monthly API credit
Core and Pro model access.
Defined allowance, no unlimited claims.
$50/month
$60 monthly API credit
Core and Pro model access.
Defined allowance, no unlimited claims.
$100/month
$130 monthly API credit
Core and Pro model access.
Defined allowance, no unlimited claims.
Proposed USD prices before taxes. Monthly credit resets each cycle. Paid access is not live yet.
Compare token rates and pay-as-you-go access →We’re building for developers who want
more say in how their tools work.
Open-weight models are the foundation. Our goal is to make model choice part of your workflow, not a decision made for you.
The model philosophyChoose a model for each request. We evaluate instruction following, coding quality and unnecessary refusals together.
The control philosophySeparate input and output rates, monthly allowances, and spending controls. Your application stays in charge of tool execution.
The API designYou bring the context. You set the scope. Your tools should help you get from intention to working code.
We’re exploring abliterated models as a foundation for that experience — and measuring whether fewer refusals translate into better work.
Understand abliterationOur research starts with a simple question:
does the agent actually finish the job?
Task completion, instruction adherence, cost, and unnecessary refusal. Read how we plan to evaluate them.
Explore the methodologyAblitron is preparing monthly subscriptions with API credit for Core and Pro abliterated models. Plans are proposed from $20/month and are not available to purchase yet.
Abliterated model access for coding and applications. We are developing Ablitron Code, starting with an OpenCode integration, alongside our developer API.
Not yet. The website describes our launch proposal. No public API, checkout or subscription is live.
Yes. Your monthly allowance pays for API requests across Core and Pro at their respective token rates.
It adds bonus API credit compared with an equal pay-as-you-go purchase. The number of tokens depends on the chosen model and the input/output mix.
We have not measured that yet. Core is the lower-cost option; speed and coding quality still need testing.
Ablitron Code is in development, starting with an OpenCode integration. A standalone Ablitron client is not available yet. The planned coding experience and direct API requests will share the same credit balance.
Explore proposed rates and allowances.
No payment or signup required.