whichllm
AI
local llm model selection
SHOW HN2026
283points
68comments
#160of 4,366 in 2026
Top 4%of 2026 by points
How HN took it
MixedNo clear lean
6 positive, 15 neutral or asking, 2 mixed, 16 critical, out of 39 comments about the project. Thread score -0.23.
Idea: 2 for, 4 againstExecution: 0 for, 3 againstPerformance: 0 for, 3 against
The thread, comment by comment
Each row is a top-level comment and its replies, in HN's order. Select a cell to open that comment on HN.
positive neutral or mixed critical about another commenter or something else the maker
What HN said
What people liked
Useful concept for picking local models by hardware
From 3 comments
Only thing I would add is the ability to point to new/uncataloged benchmarks. If I have a favorite benchmark that best matches my use-case, the ability to point to it and have it…
zambelli on HNCool idea – the local LLM space really needs a tool like this that actually understands hardware and real benchmarks instead of just “biggest model that fits.”
vaporaviatorlab on HNCool idea, thanks for making this
wald3n on HN
Objections
Thread suspects AI-generated project and promotion; maker's account looks inauthentic
From 3 comments
1. The results of this tool are not good. It’s recommending outdated models like Qwen2.5 series and missing good new models. 2. Nothing about this needs to run on your local…
Aurornis on HNThere is apparently a marketing.md file that was deleted 25min ago with the strategy to post on HN.…
hazelnut on HNOP is a newish user, all of their responses here are copied straight from Claude, and this project has an LLM slop readme (count them: 48 em-dashes on the page!) and LLM code.…
rafram on HN
Recommends outdated models such as Qwen2.5 and misses newer ones
From 2 comments
1. The results of this tool are not good. It’s recommending outdated models like Qwen2.5 series and missing good new models. 2. Nothing about this needs to run on your local…
Aurornis on HNBrew install is broken It seems pretty rubbish I have to say, its recommending me loads of qwen 2.5 which are really old and I'm easy running qwen3.5 and 3.6 models on this mac…
Bigsy on HN
Estimates only, does not measure speed on the user's machine
From 1 comment
Cool, but it looks like it doesn’t actually test anything on your machine? It does hardware detection and then some lookups. Maybe I missed it but I really want a tool like this…
pbronez on HN
Brew install reported broken
From 2 comments
Brew install is broken It seems pretty rubbish I have to say, its recommending me loads of qwen 2.5 which are really old and I'm easy running qwen3.5 and 3.6 models on this mac…
Bigsy on HNTried it on a 4060, got Qwen3-14B Q3_K_M. Matches what I actually run. Brew install failed for me too though (macOS 14.5).
Bekamakhara on HN
Unified memory and iGPU detection is wrong
From 2 comments
This doesn't correclty detect the unified memory architecture for GPU 0: STRXLGEN — 8.0 GB (ROCm 6.19.8-200.fc43.x86_64) — BW: N/A CPU: AMD RYZEN AI MAX+ 395 w/ Radeon 8060S — 16…
cyanydeez on HNit'd be nice if it had igpu support, it cant even detect it. overall great tool though. happy this exists.
justindotdev on HN
Requests
Support use-case based recommendations such as coding or video
From 2 comments
Interesting concept! A suggestion: `whichllm <USE_CASE>` would be more beneficial, i.e. `which coding` or `which text-to-video`.
armcat on HN"Best LLM" doesn't really depend on hardware alone. It actually depends more on your needs - type of workload, context length needed etc.
zkmon on HN
Support sliding-window attention in VRAM estimates
From 2 comments
The plan command is clever. How do you handle the VRAM estimation for models with sliding window attention vs full context? Something like Mistral at 32k context uses way less KV…
Jasssss on HNNice! Mistral 7B v0.1 is sliding_window: 4096 in the HuggingFace config.json (though v0.2 sets it to null). Gemma 2 alternates sliding window (4096) and full attention every…
Jasssss on HN
Add more quantization formats such as IQ3_M
From 1 comment
can you add in the other quants like IQ3_M? also my personal simple rule of thumb for local ai sizing is: max model size (GB) = ram (GB) / 1.65
macwhisperer on HN
Allow custom benchmarks
From 1 comment
Only thing I would add is the ability to point to new/uncataloged benchmarks. If I have a favorite benchmark that best matches my use-case, the ability to point to it and have it…
zambelli on HN
Questions
Where do the benchmark metrics come from
From 1 comment
Where are the metrics being sourced from exactly? Externally? or does this project have benchmarks running somewhere for the purposes of this project. Latter would likely be more…
drzaiusx11 on HN
How is the model selected, is AI used
From 1 comment
How it select model? using AI?
swaminarayan on HN
Alternatives named
llmfit
From 2 comments
What’s new regarding llmfit? https://github.com/AlexsJones/llmfit
llagerlof on HNOther than it (whichllm) being written in Python, nothing else. I just use llmfit.
rvz on HN
canirun.ai
From 2 comments
This is very helpful too: https://www.canirun.ai/
jordiburgos on HNcanirun.ai does similar, but doesn't require any installs. I've found it to be pretty accurate for my setup too.
Jeremy1026 on HN
artificialanalysis.ai
From 1 comment
Not perfect, but I find the artificialanalysis.ai "Intelligence vs. Output Tokens Used in Artificial Analysis Intelligence Index" chart[0] (scroll down to the titled chart) to be…
karmakaze on HN
RapidMLX
From 1 comment
Cool, but it looks like it doesn’t actually test anything on your machine? It does hardware detection and then some lookups. Maybe I missed it but I really want a tool like this…
pbronez on HN