🚀 Running career-ops completely free via opencode + OpenRouter — 100k+ daily tokens, renewed every 24h

[nicolasgustavoramos]

Hey everyone — wanted to share a setup I got working that lets you run career-ops at zero cost, fully in the cloud, using free AI agents.

Stack: career-ops → opencode → OpenRouter (free-tier models) → cloud AI agents
By linking opencode to OpenRouter and selecting free-tier models, I've been able to drive career-ops without spending a single cent. The free tier provides 100,000+ tokens per day, and they reset every 24 hours — more than enough for active job-search workflows.

One thing that surprised me: I'm interacting with career-ops entirely in natural language — no JSON wrangling, no config files mid-session. I just talk to it in plain Spanish (or whatever language I feel like) and it understands. That alone made the whole workflow feel way more accessible.

Still testing and iterating, but so far it's been rock solid. If I land a job using this setup, I'll come back and write up a full guide — exact model choices, token budgeting tips, and the prompts that worked best.

Stay tuned. Fingers crossed 🤞

[bracketouverte]

Are the free tier models powerful enough? Whicxh one do you recommend? How's quality?

I played with it, and Nvidia solutions, nothing as good as Claude (yet).

[nicolasgustavoramos]

Great questions! To answer them properly, it really depends on what parameters you consider important — and that's worth clarifying, because "powerful" means different things depending on the use case. For example:

• Benchmark performance (reasoning, coding, math)
• Context window size (how much the model can "remember" in one session)
• Model size (total parameters)
• Specialized strengths (coding agents, multimodal, reasoning)

Each of these gives a different picture of what's "best."

That said, I've been testing all the free tier models available on OpenRouter for my project (career-ops agent), and here's how I'd rank them by overall performance:

1. NVIDIA: Nemotron 3 Super
2. OpenAI: gpt-oss-120b
3. DeepSeek: DeepSeek V4 Flash
4. MiniMax: MiniMax M2.5
5. Poolside: Laguna M.1
6. Z.ai: GLM 4.5 Air
7. Arcee AI: Trinity Large Thinking
8. Google: Gemma 4 31B
9. OpenAI: gpt-oss-20b
10. Poolside: Laguna XS.2
11. NVIDIA: Nemotron 3 Nano Omni
12. Baidu Qianfan: CoBuddy
13. NVIDIA: Nemotron 3 Nano 30B A3B
14. NVIDIA: Nemotron Nano 9B V2

And the same goes for "quality" — it's also a multi-dimensional concept. What does quality actually mean in this context? Is it:

• Speed of search and processing?
• Ratio of jobs found vs. jobs actually worth applying to?
• The quality of conversation and interaction between the app and the user?
• Actually landing a job?
• How well it automatically generates CVs and cover letters?

All of these are valid dimensions, and the answer shifts depending on which one matters most to you.

From my experience running career-ops with the agent, all those functions are covered and the tool delivers on what it promises. The one thing no program can do is invent a job that isn't there — the market is what it is. I'm still searching, but I'm confident the right opportunity will come.

Thanks for your message and for caring — I really appreciate it. Let's keep the conversation going, this is exactly the kind of exchange worth having.