Independent. Skin in the game.
BestLLMfor is built and operated by one engineer with a continuous benchmark pipeline tracking 185+ models, not a content farm.
Mohamed Meguedmi
French engineer based in Hauts-de-France. I build independent comparison tools and writing for the open-weights AI ecosystem. Alongside BestLLMfor.com, I maintain a public LLM data API (CC BY 4.0) and an open-source MCP server for the same ecosystem.
Why this site exists
Most "best LLM" content online falls into one of two buckets: SEO pages written without ever running a model, or vendor pages selling their own benchmarks. Neither gives you what you actually need: a model that works on your hardware, fits your budget, and answers your specific use case.
BestLLMfor exists to do that work in the open: index every model worth running, benchmark them with a reproducible methodology, and ship the raw data as a free API anyone can use.
The catalog grows every week as new open-weight models ship and existing ones get re-benchmarked. Everything is versioned and dated, so you can see exactly what changed and when.
How the data is built
| Models indexed | 185+ (and counting) |
| Catalog refresh | Daily · automated pipeline |
| Quantizations tested | Q4_K_M · Q5_K_M · Q8_0 · FP16 |
| Public API | 178 JSON endpoints · CC BY 4.0 |
| MCP server | Open source on GitHub |
| Reproducibility | Model versions · quants · prompts published |
Every benchmark is documented end-to-end: model hash, quantization, runtime config, prompt set, date. See full methodology →
No funding, honest about affiliate links
No VC backing. No sponsored rankings. No paid placements. If a model leads a ranking on this site, it's because the benchmark numbers say so — that has not changed and will not. The cost calculator uses public pricing from each vendor and reflects what you'll actually pay — no rebate kickbacks.
What has changed: since June 2026, some hardware links on this site (mostly “which GPU should you buy” boxes) are Amazon Associates affiliate links — as an Amazon Associate, BestLLMfor earns from qualifying purchases, at no extra cost to you. Those links never decide which model wins a comparison; they only point to hardware capable of running the model already recommended by the benchmarks. Every page carrying one says so next to the link.
Operating costs are low (one engineer, one rig, static hosting). The project will eventually monetize further via a separate B2B consulting offer — IAPRO.fr, which sets up sovereign on-prem AI stacks for European SMEs — but BestLLMfor's editorial content and rankings remain independent and are never sponsored.