Value
Frontier coding,
engineered to be cheap.
Dropstone delivers frontier-tier capability at a fraction of what equivalent quality costs at a frontier lab. The price advantage is real engineering, not a subsidy.
4×
More usage per seat
$15/mo
Dropstone Pro
May 2026
Production at Blankline
100%
Used internally
How we got here
Blankline Research has rewritten the serving architecture around batching, scheduling, and memory layout to a degree most teams cannot justify building in-house. That is why a Dropstone Pro seat costs a fraction of what equivalent capability costs at a frontier lab. The savings come from how the system is built, not from undercharging.
Continuous batching pulls more work through the same hardware. A custom scheduler keeps the GPUs at high utilization without starving latency-sensitive requests. Memory layouts are tuned for the specific tier each request lands on, instead of treating every model as a black box. None of this is novel research; what is unusual is committing to it at this depth for a product, rather than a paper.
The visible outcome is a single number on the pricing page. The hidden outcome is that the same architecture lets us promote and demote model tiers every month without a price shock, because the cost structure does not collapse when the underlying model changes.
What we mean by value
06 principles
The price is engineering, not a subsidy.
Blankline Research rewrote the serving architecture around continuous batching, custom scheduling, and tier-aware memory layouts. The savings come from how the system is built. We are not selling under cost to acquire users, and we are not running a clock on a discount.
One price, full capability.
Vision, long-context, tool use, and multi-step planning are in-tier. There is no separate add-on for images, no surge pricing for context length, no metered fee for agentic workflows. The plan you pay for is the plan you can use end to end.
Three tiers, one honest choice.
Fast for interactive moments where the loop matters more than the answer. Pro for the everyday driver: plans, refactors, code that has to compile the first time. Heavy for the hard cases. You pick the tier that matches the work in front of you, not the other way around.
Audited every month, no quiet swaps.
Each tier is reviewed and re-selected on a monthly cycle. If a candidate model wins the review, the tier version bumps and you can see it on the model page. If nothing meaningfully improved, the version stays where it is. You will never see a quiet swap behind the same label.
What we ship runs on what we sell.
Since May 2026, Dropstone is the default coding agent across Blankline Research and Blankline product engineering. There is no separate, more expensive setup for internal use. If a tier is not good enough for our own engineers, it does not ship.
Nothing is retained, nothing is resold.
Prompts, code, and outputs are never retained, shared with the model's original operator, or routed through inference paths controlled by labs that train on customer data. The unit economics work without resorting to data as a second revenue line.
Used in-house
Since May 2026, Dropstone is the default coding agent across Blankline Research and Blankline product engineering. Every feature you see ship, including the platform you are reading this on, passed through the same Pro and Heavy tiers we sell.
That is the strongest validation we can give: we do not maintain a separate, more expensive setup for internal use. If a tier is not good enough for our own engineers, it does not ship.
Or read about how the models are selected.