Austin Ginder runs Claude Code on the $200/mo plan. He has for a while. The work is real, the model is Opus 4.7, and the bill matches the tier.
On May 4, he tried DeepSeek Pro routed through Command Code on the $1/mo plan and posted the dashboard.
What the dashboard shows
Austin didn't share a benchmark. He shared his Command Code usage screen.
- 21.5M tokens processed
- $0.75 consumed of $9.25 available on the plan
- 656 runs, all completed against
deepseek/deepseek-v4-pro
For context, his quote from the post:
"Okay I'm impressed with the DeepSeek Pro via Command Code. I can see myself offloading some of my Opus 4.7 work over to it. Also the current $1/month price is a no brainer."
The reply that mattered
In the thread, @aryanrajseo pointed out that the higher Claude tiers run $100/mo and $200/mo. Austin didn't hedge.
"Indeed! I'm a heavy $200/month Claude Code user and I can say that Command Code's handle on DeepSeek is really good. Every Claude user should try it out."
A separate reply asked how Command Code's agent harness compares to OpenCode, Codex, and Droid. Austin kept it honest:
"Not sure, I primarily use Claude Code. Command Code seems to behave how I'd expect it to."
He's not switching agents. He's not running comparison benchmarks. He's a Claude Code user who picked up a $1 plan, ran 656 jobs through it, and noticed the harness held.
Why this matters
The shift Austin describes is small and specific. Not "I'm replacing my tools." Not "DeepSeek beat Opus." Just: the Opus calls that don't need Opus can go to DeepSeek Pro on a dollar plan.
A $200/mo plan and a $1/mo plan don't compete for the same customer. They compete for the same prompt. When a heavy Claude user posts the dashboard, says the open-source-model lane is good enough, and tells other Claude users to try it, that's the kind of endorsement we can't manufacture.
Try it
1npm i -g command-codeRun cmd in your project, pick DeepSeek Pro, and watch your usage screen. The $1/mo plan covers the offloading Austin is doing.
Sign up for Command Code · Read the docs · See the original post
