HN Explorer

Ask HN: What is your AI harness that lets you switch LLM models easily?

1 points by hbarka ↗ HN
Looks like we are at the end of the Fable-level model at subscription pricing. It will be double the cost at paid-per-token usage. I’m counting on the other models to compete now at the subscription level and I need my harness to be agnostic. What do you use?

16 comments

[ 2.0 ms ] story [ 41.9 ms ] thread
zloy88
I just stay with Opus-4.8 - because it's not like that Opus is not capable of solving all my daily tasks. It just works fine. Fable is of course better, especially on paper, but it didn't do wonders for me. I feel like Anthropic can't continue just making better models all the time, they will have to care efficiency more and more because that is what they lack the most. Just compare it to any other LLM, they are all way more efficient which means they are way faster and cost less money per token. So for me it's just wait and relax - use Opus-4.8 until Anthropics models hopefully become cheaper and faster.
ricardobeat
Crush is pretty nice. It’s become my daily driver for personal projects. They maintain a list of supported providers so most of the time no configuration is needed.

OpenCode is also fine but more finnicky - if you’re ok with desktop apps, give OpenChamber a try.

sejje
What's the difference in crush and opencode? What do you mean by finicky?
PaiDxng
I’d optimize for provider-agnostic tooling now. If Fable-style subscriptions become token billing, switching between Claude, GPT, Gemini, and local models matters more than finding one perfect plan.
vinhnx
I build VT Code, a Rust coding agent, from it, you can easily switch LLM providers/models and various configs like effort and plan/build/auto mode. Fable 5 included.

> https://github.com/vinhnx/VTCode

russlan
imo model switching only works if you normalize outputs and keep model specific weirdness behind well designed adapters
mindcrash
Pi (https://pi.dev), a TUI based harness, lets you easily cycle through all configured models through the keybindings CTRL+P (forward) and SHIFT+CTRL+P (backwards).

Personally using it for switching from and to locally served Gemma and Qwen endpoints.

jarodrh
Hmm, I honestly think a harness should be more about workflow, control, ease of use, memory optimization...other things I can't think of right now, the model/subscription being the least of them. Most of my experience has come from using Claude Code so I can only speak from that, but I will at some point pick up 'Pi' when I'm feeling adventurous.

I feel completely free to use any number of subscriptions alongside Claude. Nothing restricts me from utilizing LLM API keys or CLIs pinned to roles in the harness via config. However, that being said, I use this harness confident in using Anthropic models as my main model provider alongside other models (GPT, GEMINI, Cursor - Composer 2.5 etc) as more like workhorse models to spread/manage cost on large projects.

If cost is the underlying concern, then you might want to consider your actual cost associated with each task you run on your top frontier model. It might just be that not all of your tasks need to be sent to the most expensive model.

kelrouter
I would recommend just connect your whatever you used or built to a LLM gateway & router to use any models without management complexity.

I built BitRouter(https://github.com/bitrouter/bitrouter), open-sourced, 0 markup for all models, free BYOK/subscription, and works with any harnesses out of the box