Self-hosted coding assistant vs GitHub Copilot
Copilot sends your code to GitHub's servers to generate a suggestion, then discards or retains it according to the plan you're on. A self-hosted assistant sends your code only to a model server you run, with no model provider in between. For many teams that difference decides the comparison before cost comes up; for the rest, the cost section below has the arithmetic.
What actually leaves your network
Every autocomplete request Copilot serves includes the surrounding file, and often other open files in the editor, sent to GitHub's inference endpoints to generate the completion. For an individual developer working on open-source code, that's a non-issue. For a team under an NDA, a defense contractor, a bank, or anyone working on code they can't show a third party, it's the whole issue. GitHub's own Copilot Trust Center documents what is and isn't retained by plan tier, and the details differ enough between Individual, Business and Enterprise that "Copilot doesn't train on my code" is not a statement you can make without checking which plan a given developer is actually on.
Running a coding model on a DGX Spark you serve yourself removes the plan-tier question. The model, the weights and the inference process run on one machine, and prompts never reach a model provider, so there's no retention setting to configure and no plan tier to get wrong. On a Spark you own and house, the guarantee is architectural. On a rented one it is a single, simpler contract: GPUwerk hosts the machine in EU-Central under an Article 28 DPA, keeps infrastructure admin access as any host does, and does not access workload content.
What the model itself can do
GPUwerk measured Qwen3-Coder 30B-A3B (AWQ, 4-bit) on a single 128 GB DGX Spark at 80.9 tokens per second decode for one user under vLLM, on 2 September 2026, full figures on the model page. That's comfortably above the point where a chat-style or inline-completion interface feels responsive. Because it's a mixture-of-experts model, it also holds up under load rather than collapsing: aggregate throughput reaches 3,235 tokens per second at 256 concurrent requests, up only 7.5% from the 128-concurrent figure, so a small team sharing one machine doesn't fight over latency the way it would with a dense model.
I won't put a number on how Qwen3-Coder's suggestion quality compares to Copilot's underlying model on a specific benchmark; that's a moving target on both sides, and GPUwerk hasn't run a head-to-head eval. What I can say: 80.9 tok/s single-user and a 262,144-token context window (measured prefill detail on the model page) is enough headroom to feed an agent large diffs and multi-file context without the model becoming the bottleneck. For wiring it into an actual editor or CLI agent loop, see coding agents against your own endpoint.
Cost at team scale
Copilot has several tiers with different seat and usage limits, and its prices change faster than a blog post, so check GitHub's own pricing page for the current per-seat cost. What I can compare on solid ground is the shape of the two cost curves. Copilot bills per seat per month, so cost scales linearly with headcount however much each developer uses it. A dedicated Spark bills $0.79 per hour, metered per minute, and serves the whole team at once; the more developers share it, the lower the effective cost per seat, until the Spark's throughput rather than your budget becomes the limit.
The crossover depends on team size and how many hours the machine runs. Around the clock, a 730-hour month is about $577, more than one Copilot seat. Business hours only, about 220 hours, is about $174, or roughly $17 to $22 a seat for a team of eight to ten. Set those numbers against GitHub's current published rate for your team size.
Where Copilot still wins
Copilot is zero-setup: install the extension, sign in, done. A self-hosted assistant needs a machine, a model and an inference server, which is real infrastructure work even if it's a few commands on a Spark. Copilot also benefits from continuous updates to a frontier model with no action on your part. If code privacy isn't a constraint for your team and nobody minds the per-seat cost, that convenience is worth something and there's no reason to build around it.
The decision comes down to one question: does any code your developers write need to stay away from an AI provider's servers? If yes, the setup cost of a self-hosted assistant buys you a setup where no model provider sees the code, instead of a policy read from a trust center page. If no, Copilot's convenience is hard to beat.