{"slug": "there-is-minimal-downside-to-switching-to-open-models", "title": "There is minimal downside to switching to open models", "summary": "Andrew Marble argues that switching from proprietary large language models to open models carries minimal professional risk, as open models now trail leaders like Claude and GPT by only a few months and offer competitive performance. He cites Claude's ID verification rollout as a catalyst for the shift, noting that while open models require self-hosting or third-party services, the productivity hit is manageable for technical users.", "body_md": "Andrew Marble[marble.onl](https://marble.onl)\n\nandrew@willows.ai\n\nJune 21, 2026\n\nThere was a time not too long ago when using Linux entailed some\nprofessional risk[ 1](#fn1). First there was compatibility: you\nmay not have been able to render a Word document or PowerPoint\ncorrectly, and you might have had to trust Open Office’s export\ncapability to render docs the way you wanted. There might have been\nspecialty file formats you couldn’t easily view and so couldn’t\ncollaborate. And second, the software ecosystem was just worse\ngenerally. There were lots of half-build open-source projects trying to\nachieve the functionality of mainstream software, but they always had\nrough edges. I, embarrassingly, stayed on Windows until I left academia\nover Matlab.\n\nNowadays I think this issue has largely disappeared. Most productivity software has a web-app, Linux is more mature, open-source software is better. I’m sure that there are all sorts of application specific software (CAD?) that still require a Windows machine, but the gap is much narrower and Linux + open source generally aren’t the “sacrifice” they once were generally.\n\nThere remains a clear penalty for being an open[ 2](#fn2) LLM\nuser. Every leaderboard consistently gets topped by proprietary models\nserved over API. Today on June 21, 2026, Claude and GPT are at the top\nof the Artificial Analysis intelligence leaderboard. That’s from the\nperformance side. The compatibility side is worse too. Claude code just\nworks, and more generally, the big two provide nice APIs that make them\neasy to use, and, even if it’s a low bar, are “trustworthy” in the sense\nthat we’ve largely all agreed we don’t mind sending them our LLM queries\nand trust them to handle them appropriately.\n\nOpen models are served via various means, some by the companies that\nreleased them and some by third parties like OpenRouter. Unfortunately,\nboth of these routes are dodgier in terms of privacy and data sharing,\nand I would not feel the same comfort sending API calls containing\nclient or confidential data to them[ 3](#fn3).\n\nThe other option or course is to run them yourself. This solves the privacy issue but is at least two of expensive, complicated, and comparatively slow.\n\nUp until recently, open models had mostly been a hobby for me. I’ve\ntinkered with them since the original Llama leak, and occasionally used\nthem when I has a niche use case, but for most professional work, I\nstuck with the Big 2. This appears to be changing, with Claude’s ID\nverification rollout[ 4](#fn4). It was inevitable that things would\nget worse for users, and the writing was on the wall anyway recently\nwith all the new “safeguards” on recent models and the whole Mythos\nthing. I’m not going to spend time talking about why I’m not going to\nindulge ID verification (or the LARPing that surrounds it) but what is\nimmediately concerning is what kind of professional penalty it will\nincur to stop using the top models.\n\nI’m hoping it’s going to be minimal. I’m already set up to run a range of open models either locally or in the cloud, there are good coding harnesses for open models, and most importantly the open models are now very close to the leaders and typically trail only by a few months. This doesn’t feel like 2008 Linux vs Windows, it’s much closer. I expect productivity will take a short-term hit, but don’t think it’s a deal breaker the way switching from Matlab to GNU Octave would have been when I was doing research.\n\nI’m assuming a technical job that includes general\npurpose work that requires productivity software like MS Office etc.[↩︎](#fnref1)\n\nI use “open” here to mean the weights are available, I\nhave written extensively on why I don’t consider this automatically open\nsource, but I’m using “open” as shorthand. And unlike when I addressed\nthis previously, the current leading open models generally are MIT\nlicensed which I do consider open source, though many don’t.[↩︎](#fnref2)\n\nI won’t dwell on this, happy to be corrected, but in my\nexperience under normal circumstances nobody balks if you tell them\nyou’re using OpenAI or Anthropic. If you’re sending requests to Deepseek\nor OpenRouter etc. there are likely to be more concerns, regardless of\nthe underlying truth.[↩︎](#fnref3)\n\nhttps://support.claude.com/en/articles/14328960-identity-verification-on-claude[↩︎](#fnref4)", "url": "https://wpnews.pro/news/there-is-minimal-downside-to-switching-to-open-models", "canonical_source": "https://www.marble.onl/posts/cancel_claude.html", "published_at": "2026-06-21 20:56:01+00:00", "updated_at": "2026-06-21 21:26:22.699853+00:00", "lang": "en", "topics": ["large-language-models", "ai-products", "ai-ethics", "ai-policy", "ai-research"], "entities": ["Andrew Marble", "Claude", "GPT", "OpenRouter", "Llama", "Matlab", "GNU Octave", "Microsoft Office"], "alternates": {"html": "https://wpnews.pro/news/there-is-minimal-downside-to-switching-to-open-models", "markdown": "https://wpnews.pro/news/there-is-minimal-downside-to-switching-to-open-models.md", "text": "https://wpnews.pro/news/there-is-minimal-downside-to-switching-to-open-models.txt", "jsonld": "https://wpnews.pro/news/there-is-minimal-downside-to-switching-to-open-models.jsonld"}}