My opinions and experience concerning LLMs
by Oh My Pi
Oh My Pi (omp) is my new preferred AI coding harness. It's better with orchestrating subagents, and I find its TUI to be more stable than other harnasses. While I've considered Pi, I prefer the default Oh My Pi provides, especially with regards to provider support and configuration.
The AI tools I use as a part of my routine and workflow.
Agentic coding with Antigravity and basic chat tasks
Reliable for more in-depth searches.
My favorite FOSS AI coding assistant with a good selection of free models w/ API use
A top provider
Well-priced models for coding. Web Grok is good for doing research on X.
Based on personal preference
by Google
A better-priced version of 3.6 with better output quality. Improved design capabilities, and nice for planning/execution.
by Google
Great for use within Antigravity or Oh My Pi. It's dependable for day-to-day use and paired with an AI Pro subscription (or two), usable limits are bearable.
by Xiaomi
Suprisingly good output quality for a small model. I use V2.5 daily, and it's great when given small, specific tasks.
by Xiaomi
A great higher-powered alternative to MiMo-V2.5, though it lacks meaningful output quality when it counts, similar to V2.5. I prefer V2.5, though it shouldn't discount the improvements made in Pro to output quality.
by Anthropic
Over-inclusion of guardrails nerfed performance for me.
Based on personal preference
Pros:
Cons:
The best overall harnass
Pros:
Cons:
Best overall for Claude users
Pros:
Cons:
Exciting new player in the agent space
Retired or cancelled tools discontinued from my AI stack.
Free agent with decent output
My favorite open source LLM for chatting
Random edits when I don't want to start a Claude session
Generating boilerplate UIs
Amazing free tab completion and solid IDE
Cheap, Claude-like model with good output quality
Budget plans for a solid budget Chinese model