OpenCode vs Claude Code vs Copilot vs Gemini: Very Simple Review

Quick Summary

This is my hands-on very simple comparison of Gemini CLI, Copilot CLI, Claude Code, and OpenCode after months of real usage.

This is based on my personal experience, not a benchmark.

Why I tested these tools

NGL, I am a bit of a terminal nerd (I am a Neovim user btw) and I love trying new tools that can make my development workflow faster.

When I first heard about these CLIs, I was really excited to see how they would perform in real daily work, and of course see for myself what these tools could do, instead of being an AI doomer or getting caught in AI hype.

The SUT

Tools:

• Gemini CLI
• Copilot CLI
• Claude Code CLI
• OpenCode CLI

Time spent in order of usage:

• Gemini CLI: ~3 months
• Copilot CLI: 1.5 months
• Claude Code CLI: 1 month
• OpenCode CLI: 1.5 months

What I care about

• Fast (response time, and overall speed in my workflow)
• Performance (CPU and memory usage, quality of the output, etc...)
• Number of providers and integrations available
• Simple
• Reliable
• Open to the community

Tool by tool

Gemini CLI

The worst CLI of all for me. To be honest, I started with Google because my company was paying for Gemini Pro, so I ended up using it for a few months, but I never really felt I could trust it for daily work. The experience felt broken, with random HTTP errors, unclear token limit feedback, and a slow and clunky UI. The worst part was waiting several seconds for the Gemini model to answer, only to discover that for some random reason it was not available, and then I had to switch to a mini or older model just to make it work.

What did not work for me:

• Token limit feedback felt unclear
• Random HTTP errors happened too often
• Slower feel in daily usage
• UI responsiveness felt rough
• Some sessions started looping and output quality dropped
• 429 HTTP errors were so annoying

Copilot CLI

Copilot, most of the time, worked well and was a good assistant in the terminal, but it did not feel like a game changer for me. It felt more like a nice-to-have.

What did not work for me:

The monthly limits. I hit the limits multiple times and it was really frustrating, especially when you are in the middle of a workflow and suddenly cannot use the tool anymore. Compared with OpenAI and Anthropic, the limits feel more "generous" and reset every 5 hours.

Claude Code CLI

Overall the tool is good, and when you come from Copilot CLI or Gemini CLI (🪦) it feels like a big step up in capabilities and quality. But it did not fit my workflow and had too many bugs. It froze a lot and slowed down my computer by eating memory and CPU.

What worked:

• PLAN mode that lets you set a plan and execute it with the assistant (I really like this approach)
• Limits reset every 5 hours (sooo good)
• A lot of integrations with third-party tools out of the box, such as GitHub, Slack, etc.

What did not work for me:

• Felt clunky in my workflow
• UI errors were annoying (elements vanishing on scroll, or not rendering properly, should we blame React for that?)
• Hangs and freezes happened sometimes, especially in longer workflows
• CPU and memory usage can get crazy sometimes

OpenCode

OpenCode was the latest tool I tested and it was a really pleasant surprise. It felt fast and responsive, the UI is simple and clean, and the open source and community-friendly direction is really nice. I also like that it has many integrations with other tools and services. It feels like a tool built for me.

What worked:

• Speed
• Simplicity
• Open source and community-friendly direction
• Very easy way to see sessions and share it with other users
• Local model support
• Crazy amount of providers that you can connect. you can easily connect your subscriptions from Z.ai, Zen.ai, OpenAI, Anthropic, GitHub Copilot, and more.

What did not work for me:

• Missing plugins to show quota limits for each provider, such a boomer
• Some integrations are not available out of the box compared with Claude Code.

Final take

For my workflow, OpenCode was the easy winner.
Claude Code was usable but clunky for me.
Copilot CLI was good but the monthly limits were a deal breaker for me.
Gemini CLI was the hardest to trust in daily use.