Low-latency agentic coding that is no longer just a latency tier. Built on DeepSeek V4 Flash 0731, hosted in the US.
Fast moves to DeepSeek V4 Flash 0731, a retrain at unchanged architecture and price. It scores 50.0 on the independently measured AA Intelligence Index against 40.3 for the April 2026 model, the largest single move in 1.7. We no longer report Fast as a latency-only tier.
Every cycle we re-evaluate the open-weight leaderboard. When something better ships, Fast gets swapped. Same CLI, same security model.
Dropstone Fast is the tier you reach for thousands of times a day: inline completion, file-scoped edits, quick refactors, test stubs. At 1.7 it runs DeepSeek V4 Flash 0731, a retrain of the V4 Flash we shipped at 1.6 with identical architecture and identical price (284B total, 13B active, 1M context), and it scores 50.0 on the independently measured Artificial Analysis Intelligence Index against 40.3 for its predecessor. That is the largest tier move in this release: Fast now sits 1.1 points behind our Pro tier and 5.7 ahead of DeepSeek V4 Pro, a model one tier above it in its own family. At 1.6 we described Fast as a latency-and-cost tier and made no capability claim for it; at 50.0 that framing no longer holds. Fast is text-only, so image requests route to Kimi K2.7 Code, the registered vision backend, on every plan. Same approval-gated CLI, same US-hosted inference.
Low-latency suggestions that respect your repo conventions. Designed for thousand-keystroke workflows.
Rename, refactor, extract. Fast enough to feel like an IDE primitive, not a network round-trip.
Generate unit-test skeletons against the function under the cursor with the same approval gate as any other tool call.
Every Dropstone request is treated as if the model could be adversarial. The CLI requires explicit user approval before any action that writes to disk, runs a shell command, or fetches a URL. No model output is ever auto-executed.
Fast runs on SOC 2-certified, US-based inference providers. Enterprise and commercial data is never stored and never trained on, regardless of any setting. Consumer sessions help improve the open model we run and can be turned off anytime in settings, as text only, never your raw images.
Fast is built on DeepSeek V4 Flash 0731, an open-weight foundation model. Goldwasser et al. (2022) proved no party can prove a closed foundation model is free of embedded behaviors, including Anthropic for Claude and OpenAI for GPT. We say this out loud. The runtime is why model origin does not matter for your code.
Same approval gate, same US-hosted inference, same zero-retention guarantee across all three. Pick the smallest model that meets the task.
Low-latency agentic coding that is no longer just a latency tier. Built on DeepSeek V4 Flash 0731, hosted in the US.
The everyday workhorse, and the most substantive change in 1.7: last cycle's Heavy tier is now the default on every plan. Built on GLM-5.2, hosted in the US.
The strongest open-weight model measured: first open model past a flagship closed model on the aggregate index, and #1 on the Frontend Code Arena. Built on Kimi K3, hosted in the US.
Fast and Pro are included on every plan, including free. Heavy is unlocked on the Pro plan ($20/mo) and Max plan (from $100/mo).
Install the CLI, authenticate, and start running approval-gated agentic workflows. No credit card to start.
