--- title: IBM CEO says open weight models are now irresistibly... description: IBM CEO Arvind Krishna says open weight AI models are now good enough and cheap enough to replace frontier closed models for most enterprise tasks. He says the... canonical: https://daily.dev/posts/ibm-ceo-says-open-weight-models-are-now-irresistibly-good-and-cheap-on-prem-ik2jnufrl twitter:card: summary_large_image twitter:site: @dailydotdev og:type: website og:site_name: daily.dev og:title: IBM CEO says open weight models are now irresistibly good and cheap on-prem | daily.dev og:description: IBM CEO Arvind Krishna says open weight AI models are now good enough and cheap enough to replace frontier closed models for most enterprise tasks. He says the... og:url: https://daily.dev/posts/ibm-ceo-says-open-weight-models-are-now-irresistibly-good-and-cheap-on-prem-ik2jnufrl og:image: https://api.daily.dev/og/posts/IK2JnuFrL.png og:image:alt: IBM CEO says open weight models are now irresistibly good and cheap on-prem og:image:width: 1200 og:image:height: 630 og:locale: en --- > ## Documentation Index > Fetch the complete documentation index at: https://daily.dev/llms.txt > Use this file to discover all available pages before exploring further. # IBM CEO says open weight models are now irresistibly good and cheap on-prem **[Hacker News](https://daily.dev/sources/hn)** · 4 min read · 0 upvotes · 0 comments ## Summary IBM CEO Arvind Krishna says open weight AI models are now good enough and cheap enough to replace frontier closed models for most enterprise tasks. He says the performance gap narrowed from 70-to-20/30 in 2024 to 70-to-50 in late 2025, while open models now cost 10 to 25 percent of frontier token prices for about 90 percent of the effectiveness. Krishna expects most businesses can run useful on-prem AI in just two or three racks, boosting demand for IBM's Spyre mainframe accelerator cards. IBM's Q3 revenue hit $17.1 billion, up five percent, with software growing nine percent to $7.8 billion. ## Full article daily.dev links to this article rather than hosting it. Read it at the original source: ## Questions this post answers ### How much cheaper are open weight AI models compared to frontier closed models in terms of token cost? Open weight models cost roughly 10 to 25 percent of the token cost of frontier closed models, according to IBM CEO Arvind Krishna. He stated they now deliver about 90 percent of the effectiveness of top frontier models, making them an attractive option for most enterprise tasks like summarizing emails, examining contracts, and answering customer calls. _Teams weighing open weight versus frontier models for production workloads can track real-world comparisons on daily.dev._ ### How has the performance gap between open weight and closed frontier AI models changed recently? The gap narrowed significantly between 2024 and late 2025. In 2024, a top frontier model might score 70 on a broad benchmark while an open model scored only 20 to 30; by late 2025, the open model's score rose to about 50 against the same 70, according to IBM CEO Arvind Krishna. _Anyone deciding which model tier to standardize on can follow benchmark shifts like this on daily.dev._ ### How many server racks does a business need to run useful on-prem AI infrastructure? IBM CEO Arvind Krishna said most businesses can build useful AI infrastructure in just two or three racks. He noted the bigger challenge for enterprises is staffing: finding kernel engineers and network specialists who can cluster GPUs with proper network proximity and caching to get fast responses. _Engineers planning on-prem AI rollouts can keep up with infrastructure sizing guidance on daily.dev._ ## Community take How the wider developer community reacted, aggregated from 2 discussions and 117 comments across hackernews (as of 2026-10-26). **TL;DR:** Commenters broadly agree open-weight models are already good enough for serious self-hosted use, with several sharing real local-inference setups, though there's real debate over whether on-prem is actually cheaper than subsidized frontier subscriptions and skepticism toward IBM's own credibility and hardware ambitions. **Sentiment:** 45% positive · 35% mixed · 20% skeptical **The case for** - Several people report running open models locally for months and finding them good enough for agentic coding, with freedom from provider restrictions, rate limits, and content filtering as a major draw. - Keeping sensitive/proprietary data off third-party APIs is seen as a decisive advantage regardless of raw model quality. - Inference prices are expected to keep falling, and some argue local hardware can be resold or amortized, making the economics more favorable over time. - IBM's Granite models are noted as genuinely useful and free, giving the claim some independent credibility. **The pushback** - Several argue that frontier model subscriptions are VC-subsidized and that true cost comparisons (vs. API pricing) make local hosting look less cheap than claimed. - Running serious local setups still requires expensive hardware, nontrivial electricity costs, and ongoing tuning/inference engineering expertise. - Strong skepticism toward taking a CEO's self-interested claims at face value, especially given IBM's own reputation for overpriced, dysfunctional offerings. - Doubts that IBM's Spyre/Power hardware can compete with Nvidia's ecosystem (CUDA, libraries, upgrade path) any time soon. **By community** - hackernews (mixed): A lively, detailed back-and-forth where enthusiasts share real local-inference setups and champion open models for privacy and freedom, countered by skepticism about true cost savings, IBM's credibility, and whether its hardware can ever rival Nvidia. **Hottest debate:** Whether self-hosting open models is actually cheaper than frontier subscriptions once hardware, electricity, and engineering time are accounted for, versus the view that labs are merely subsidizing prices that will eventually rise. **Open questions** - Will frontier labs' subsidized subscription pricing hold up once they need to turn a profit, and how will that change the cost calculus versus self-hosting? - Can IBM's Spyre and Power hardware realistically close the ecosystem gap with Nvidia given CUDA's entrenched advantage? - Will China continue releasing competitive open-weight models, or could geopolitical/regulatory shifts cut off that supply? **Highlights** > My local model never says no. I never get banned. If I want to do security stuff, reverse engine, there it goes! Do I wanna chat about politics? No problem! Do I want it to analyze medical stuff? I don't get to hear that it's not a doctor and I should call 911. Do I want to explore legal docs, no problem! There are lots of things that cloud models will say no to! I read that don't even agree to help with advanced biology and chemistry topic much. I also never run out of tokens, no 5 hr limit. I can run it 24/7. I don't run out of tokens. I can run 16 sub agents at the same time, non stop! 1-4 billion tokens a month? No problem. Any agent harness I want! No restrictions. Privacy all day. My own true system prompt, not something someone has determined, no filters. Furthermore, there are many open weight cloud models that don't have equivalent online. For a while before cloud models got good at math, the best model at math was DeepSeekv2 Math and there was no API. The only way to use it was to run local. All sorts of audio models, image/video models, etc. No multiple subscription needed. Private UI for my family, kids included (I can guard what kids infer). Got a friend that really needs some tokens? I give em access through private VPN. Telling someone that they are paying for decades of Claude for local AI is like telling someone they are paying for decades of public transport when they buy a car. Sure, you are right, but ... > — [segmondy on hackernews · 2 comments](https://news.ycombinator.com/item?id=50239362) > It's absolutely true. I have three different setups here, a rig with 8 gpus that I use in either one of two modes (all 8, very heavy model or just one much lighter) and two DGX sparks. I haven't used a paid model in months. Sure, they use a few more tokens and the results may be slightly less good (but this doesn't really matter because I'm still in the loop so it is easy to correct). But the fact that I'm not sharing my work with the paid providers is much more to my liking than having big brother looking over my shoulder at every step and occasionally giving me the 'I can't let you do that, Jacques' treatment. > — [jacquesm on hackernews · 6 comments](https://news.ycombinator.com/item?id=50238737) > Working on a blog post to document it all, but that will take a few more days. The short version: the big rig is probably $25K, ttft is dependent on how much prefill you have but typically a couple of seconds (2K prefill/sec in one setup, about 500 in the other), tgs 100 to 150 or so depending on what it is doing and how well the speculative decoder works with the current context. Electricity is pretty bad, full tilt it's about 2500W or so but I've power limited it to 1500 to keep it civil. It is not cheaper than using a frontier lab (or at least, I don't think it is), but the independence bit is the important part for me and I don't care if that carries a surcharge, even if it is a substantial one. > — [jacquesm on hackernews · 3 comments](https://news.ycombinator.com/item?id=50238816) > Of course IBM - the guys that sell you mainframes - would be saying this. This is a win though. But not for the reasons people might think it is. It is a win because it's a massive player now countering the new (temporary) behemoths. With all the leverage, budget and connections. And contrary to the AI Labs, IBM isn't just that one SV rockstar dude, but a whole org with tons of people. Meaning that they're going to be far closer to the baseline of sane humans just from averaging alone. > — [hypfer on hackernews · 1 comments](https://news.ycombinator.com/item?id=50239573) > Can we please stop giving weight to CEOs talking their book? Like of course IBM wants you to think this, it would be great for them if it’s true. And maybe it is true! But IBM’s CEO saying it’s true tells me just as much as if a frontier lab said it’s not: nothing. > — [collinmcnulty on hackernews · 4 comments](https://news.ycombinator.com/item?id=50238658) **Source threads** - [hackernews](https://news.ycombinator.com/item?id=50191944) · 16 points · 14 comments - [hackernews](https://news.ycombinator.com/item?id=50238088) · 178 points · 103 comments ## Similar posts on daily.dev - [Mistral Says 99% of Enterprise AI Can Be Open. That Should Terrify the Closed-Model Business](https://daily.dev/posts/mistral-says-99-of-enterprise-ai-can-be-open-that-should-terrify-the-closed-model-business-eq4tuzn9x) · Medium · 1 upvotes · 0 comments - [IBM CEO Arvind Krishna says there is no AI bubble after all](https://daily.dev/posts/ibm-ceo-arvind-krishna-says-there-is-no-ai-bubble-after-all-0jxi2yrpg) · The Verge · 1 upvotes · 0 comments - [The State of Open Source AI — V1.0 · July 2026](https://daily.dev/posts/the-state-of-open-source-ai-v1-0-july-2026-gjzwgazbw) · Hacker News · 1 upvotes · 0 comments --- Tags: [#open-source](https://daily.dev/tags/open-source), [#llm](https://daily.dev/tags/llm), [#infrastructure](https://daily.dev/tags/infrastructure), [#ibm](https://daily.dev/tags/ibm) [View this post on daily.dev](https://daily.dev/posts/ibm-ceo-says-open-weight-models-are-now-irresistibly-good-and-cheap-on-prem-ik2jnufrl) ```json {"@context":"https://schema.org","@graph":[{"@type":"Organization","@id":"https://daily.dev/#organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180},"sameAs":["https://twitter.com/dailydotdev","https://github.com/dailydotdev","https://www.linkedin.com/company/daily-dev-ltd"]},{"@type":"WebSite","@id":"https://daily.dev/#website","url":"https://daily.dev","name":"daily.dev","publisher":{"@id":"https://daily.dev/#organization"},"potentialAction":{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https://daily.dev/search?q={search_term_string}"},"query-input":"required name=search_term_string"}}]} {"@context":"https://schema.org","@type":"TechArticle","headline":"IBM CEO says open weight models are now irresistibly good and cheap on-prem","url":"https://daily.dev/posts/ibm-ceo-says-open-weight-models-are-now-irresistibly-good-and-cheap-on-prem-ik2jnufrl","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/ibm-ceo-says-open-weight-models-are-now-irresistibly-good-and-cheap-on-prem-ik2jnufrl"},"datePublished":"2026-10-26T03:25:05.407Z","dateModified":"2026-10-26T07:36:56.701Z","description":"IBM CEO Arvind Krishna says open weight AI models are now good enough and cheap enough to replace frontier closed models for most enterprise tasks. He says the...","image":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/412448f856d0653cd2cac68c0777d773?_a=AQAEuop","thumbnailUrl":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/412448f856d0653cd2cac68c0777d773?_a=AQAEuop","isAccessibleForFree":true,"articleSection":"Hacker News","inLanguage":"en","publisher":{"@type":"Organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180}},"author":{"@type":"Organization","name":"Hacker News","logo":"https://media.daily.dev/image/upload/t_logo,f_auto/v1/logos/hn","url":"https://daily.dev/sources/hn"},"commentCount":0,"discussionUrl":"https://daily.dev/posts/ibm-ceo-says-open-weight-models-are-now-irresistibly-good-and-cheap-on-prem-ik2jnufrl","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":0},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":0}],"keywords":"open-source,llm,infrastructure,ibm","timeRequired":"PT4M"} {"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"Hacker News","item":"https://daily.dev/sources/hn"},{"@type":"ListItem","position":3,"name":"IBM CEO says open weight models are now irresistibly good and cheap on-prem"}]} {"@context":"https://schema.org","@type":"FAQPage","@id":"https://daily.dev/posts/ibm-ceo-says-open-weight-models-are-now-irresistibly-good-and-cheap-on-prem-ik2jnufrl#faq","mainEntity":[{"@type":"Question","name":"How much cheaper are open weight AI models compared to frontier closed models in terms of token cost?","acceptedAnswer":{"@type":"Answer","text":"Open weight models cost roughly 10 to 25 percent of the token cost of frontier closed models, according to IBM CEO Arvind Krishna. He stated they now deliver about 90 percent of the effectiveness of top frontier models, making them an attractive option for most enterprise tasks like summarizing emails, examining contracts, and answering customer calls. Teams weighing open weight versus frontier models for production workloads can track real-world comparisons on daily.dev."}},{"@type":"Question","name":"How has the performance gap between open weight and closed frontier AI models changed recently?","acceptedAnswer":{"@type":"Answer","text":"The gap narrowed significantly between 2024 and late 2025. In 2024, a top frontier model might score 70 on a broad benchmark while an open model scored only 20 to 30; by late 2025, the open model's score rose to about 50 against the same 70, according to IBM CEO Arvind Krishna. Anyone deciding which model tier to standardize on can follow benchmark shifts like this on daily.dev."}},{"@type":"Question","name":"How many server racks does a business need to run useful on-prem AI infrastructure?","acceptedAnswer":{"@type":"Answer","text":"IBM CEO Arvind Krishna said most businesses can build useful AI infrastructure in just two or three racks. He noted the bigger challenge for enterprises is staffing: finding kernel engineers and network specialists who can cluster GPUs with proper network proximity and caching to get fast responses. Engineers planning on-prem AI rollouts can keep up with infrastructure sizing guidance on daily.dev."}}]} ```