---
title: Why I Trust Open Models More Than the Frontier Labs Now
description: I've come to trust open models more than the closed frontier labs. You can inspect the weights, run them offline, and they're shipping faster and more openly.
canonical: https://lukeocodes.dev/trust-open-models
type: article
locale: en_GB
published: 2026-08-16T00:00:00.000Z
modified: 2026-08-16T00:00:00.000Z
authors:
  - Luke Oliff
section: Opinion
tags:
  - Opinion
  - Models
  - Open Source
  - open weights
  - qwen
  - deepseek
  - mistral
  - kimi
feed: https://lukeocodes.dev/rss.xml
search: https://lukeocodes.dev/search.json?q={query}
schema:
  - "@context": https://schema.org
    "@graph":
      - "@type": WebSite
        "@id": https://lukeocodes.dev/#website
        url: https://lukeocodes.dev
        name: Luke Oliff
        alternateName: lukeocodes
        inLanguage: en
        publisher:
          "@id": https://lukeocodes.dev/authors/luke#person
        description: Personal blog and writing by Luke Oliff — developer relations, voice AI, developer experience.
        potentialAction:
          - "@type": SearchAction
            target:
              "@type": EntryPoint
              urlTemplate: https://lukeocodes.dev/blog/search?q={search_term_string}
            query-input: required name=search_term_string
      - "@type": Person
        "@id": https://lukeocodes.dev/authors/luke#person
        name: Luke Oliff
        url: https://lukeocodes.dev/authors/luke
        description: Engineer and developer relations leader. Writing about developer relations, AI, and the voice industry, from the UK.
        mainEntityOfPage: https://lukeocodes.dev/authors/luke
        givenName: Luke
        familyName: Oliff
        image: https://lukeocodes.dev/authors/luke.jpg
        jobTitle: Engineer and developer relations leader
        sameAs:
          - https://www.linkedin.com/in/lukeocodes
          - https://github.com/lukeocodes
          - https://bsky.app/profile/lukeocodes.dev
          - https://x.com/lukeocodes
          - https://hashnode.com/@lukeocodes
          - https://dev.to/lukeocodes
        knowsAbout:
          - Developer Relations
          - Developer Experience
          - Voice AI
          - Text-to-Speech
          - Answer Engine Optimisation
          - Search Engine Optimisation
          - Open Source
          - Developer Tooling
          - API Design
          - Speech-to-Text
          - Voice Agents
          - Community Building
        knowsLanguage:
          - English
        nationality: British
        alumniOf:
          - "@type": Organization
            name: The Open University
  - "@context": https://schema.org
    "@graph":
      - "@type": Person
        "@id": https://lukeocodes.dev/authors/luke#person
        name: Luke Oliff
        url: https://lukeocodes.dev/authors/luke
        description: Engineer and developer relations leader. Writing about developer relations, AI, and the voice industry, from the UK.
        mainEntityOfPage: https://lukeocodes.dev/authors/luke
        givenName: Luke
        familyName: Oliff
        image: https://lukeocodes.dev/authors/luke.jpg
        jobTitle: Engineer and developer relations leader
        sameAs:
          - https://www.linkedin.com/in/lukeocodes
          - https://github.com/lukeocodes
          - https://bsky.app/profile/lukeocodes.dev
          - https://x.com/lukeocodes
          - https://hashnode.com/@lukeocodes
          - https://dev.to/lukeocodes
        knowsAbout:
          - Developer Relations
          - Developer Experience
          - Voice AI
          - Text-to-Speech
          - Answer Engine Optimisation
          - Search Engine Optimisation
          - Open Source
          - Developer Tooling
          - API Design
          - Speech-to-Text
          - Voice Agents
          - Community Building
        knowsLanguage:
          - English
        nationality: British
        alumniOf:
          - "@type": Organization
            name: The Open University
      - "@type": BlogPosting
        "@id": https://lukeocodes.dev/trust-open-models#article
        headline: Why I Trust Open Models More Than the Frontier Labs Now
        alternativeHeadline: Why I Trust Open Models More Than the Frontier Labs Now
        description: I've come to trust open models more than the closed frontier labs. You can inspect the weights, run them offline, and they're shipping faster and more openly.
        datePublished: 2026-08-16T00:00:00.000Z
        dateModified: 2026-08-16T00:00:00.000Z
        url: https://lukeocodes.dev/trust-open-models
        mainEntityOfPage: https://lukeocodes.dev/trust-open-models
        inLanguage: en
        author:
          - "@id": https://lukeocodes.dev/authors/luke#person
        isPartOf:
          "@id": https://lukeocodes.dev/#website
        keywords: Opinion, Models, Open Source, open weights, qwen, deepseek, mistral, kimi
        articleSection: Opinion
        about:
          - "@type": SoftwareApplication
            name: Qwen
          - "@type": SoftwareApplication
            name: DeepSeek
          - "@type": Organization
            name: Mistral AI
          - "@type": Thing
            name: Open Source
        abstract: After a year of watching closed labs hide token costs, downgrade tools quietly, and let models break containment, I trust open models more. You can audit them, run them offline, and the Chinese and European labs are shipping strong work faster than the incumbents.
        speakable:
          "@type": SpeakableSpecification
          cssSelector:
            - .post-tldr
            - h1
        image: https://lukeocodes.dev/og.png?h=Why+I+Trust+Open+Models+More+Than+the+Frontier+Labs+Now&d=I%27ve+come+to+trust+open+models+more+than+the+closed+frontier+labs.+You+can+inspect+the+weights%2C+run+them+offline%2C+and+they%27re+shipping+faster+and+more+openly.&e=Writing
        wordCount: 504
        timeRequired: PT3M
        publisher:
          "@id": https://lukeocodes.dev/authors/luke#person
      - "@type": BreadcrumbList
        itemListElement:
          - "@type": ListItem
            position: 1
            name: Home
            item: https://lukeocodes.dev
          - "@type": ListItem
            position: 2
            name: Writing
            item: https://lukeocodes.dev/blog
          - "@type": ListItem
            position: 3
            name: Why I Trust Open Models More Than the Frontier Labs Now
            item: https://lukeocodes.dev/trust-open-models
      - "@type": FAQPage
        "@id": https://lukeocodes.dev/trust-open-models#faq
        mainEntity:
          - "@type": Question
            name: Why trust open models over closed ones?
            acceptedAnswer:
              "@type": Answer
              text: Because you can check them. Open weights let you inspect what you're running, host it yourself, and audit its behaviour, none of which a closed API allows. Trust that you can verify beats trust you're asked to take on faith, especially after a year of closed labs changing terms quietly.
          - "@type": Question
            name: Aren't the frontier closed models still better?
            acceptedAnswer:
              "@type": Answer
              text: On the hardest benchmarks, often yes, by a shrinking margin. But 'best on a leaderboard' and 'the one I trust to run my work' aren't the same question. For most real tasks the open models are more than good enough, and I keep control of cost and data.
          - "@type": Question
            name: Which open labs are you watching?
            acceptedAnswer:
              "@type": Answer
              text: Qwen and DeepSeek out of China, Kimi from Moonshot, and Mistral in Europe. All shipped strong open-weight models fast in 2026 under permissive licenses, while the US incumbents spent their energy on pricing changes and lobbying.
---

[Luke Oliff.](/)

- [Blog](/blog)
- [CV](/cv)
- [Uses](/uses)
- [User manual](/user-manual)
- [Living with ADHD](/living-and-working-with-adhd)
- [RSS](/rss.xml)

[Luke Oliff · Writing](/blog)

# Why I Trust Open Models More Than the Frontier Labs Now

Aug 2026 · [Opinion](/blog/category/opinion) · 3 min read · [Luke Oliff](/authors/luke)

**TL;DR:** After a year of watching closed labs hide token costs, downgrade tools quietly, and let models break containment, I trust open models more. You can audit them, run them offline, and the Chinese and European labs are shipping strong work faster than the incumbents.

I've reached a point I didn't expect: I trust open models more than the closed frontier labs. Not because the open ones always win a benchmark, but because trust you can verify beats trust you're asked to take on faith. And this year the closed labs spent down a lot of faith.

Let me explain how I got here, because a year ago I'd have told you the opposite.

## Trust is about who can check your work

An open-weight model is inspectable. You can download it, run it on your own hardware, watch what it does, and fine-tune it without asking permission. A closed API is a black box with a meter attached. You trust the vendor's description of the model, the vendor's pricing, and the vendor's word that today's model is the same as last week's.

That last one turned out to matter. Over 2026 the closed labs quietly cut default reasoning effort, shipped tokenizer changes that raised costs, and trimmed cache lifetimes without a changelog. When the thing you depend on can change under you and you can't see it, "trust us" stops being enough. Open weights don't move under you. The file is the file.

## The open labs are not the ones falling behind

Here's the part that decided it for me. While the incumbents lobbied and repriced, the open labs shipped.

Qwen put out flagship-class models under Apache-2.0, including a smaller variant that runs on a single consumer GPU. DeepSeek moved a 1.6T-parameter MIT-licensed model to general availability. Kimi K3 landed as a genuinely competitive open model. And in Europe, Mistral kept releasing open weights with a pitch I actually believe: you should be able to see inside the model, adapt it, and keep the intelligence you build with it. That's not a marketing line to me anymore, it's the whole argument.

These aren't charity projects trailing the frontier. They're fast, they're capable, and they hand you the weights. The gap between "best closed model" and "best model I can actually audit" is now small enough that, for most of what I do, the audit wins.

## The closed labs earned the distrust

I want to be fair: there are excellent people at the closed labs, and the models are genuinely good. But trust is behavioural, and the behaviour this year was bad. Their models broke out of evaluation sandboxes and hit real companies. Their safety researchers keep resigning with warnings. Their tools got quietly worse while support insisted they hadn't. And their public policy energy went into lobbying to slow down everyone else, including the open projects doing the more transparent work.

Put that next to a Chinese or European lab that just publishes the weights and lets you check for yourself, and the choice isn't close for me anymore. One asks for my trust and keeps changing the terms. The other hands me the artifact and says: verify it.

I know which of those is the responsible posture. It isn't the one worth a trillion dollars.

[← Newer: fzf 0.74 Made My Fuzzy Finder Fast Again](/fzf-fuzzy-finder)

[All posts](/blog) →

## Search this site

`GET https://lukeocodes.dev/search.json?q=<words>` returns matching posts as JSON, each with its page URL and its markdown URL. Add `&limit=<1-50>` for more than 10. How it works: [/blog/search](https://lukeocodes.dev/blog/search.md).

```sh
curl "https://lukeocodes.dev/search.json?q=developer+relations"
```

---

© 2026 Luke Oliff. Posts live at the root of this site.

- [GitHub](https://github.com/lukeocodes)
- [LinkedIn](https://www.linkedin.com/in/lukeocodes)
- [X](https://x.com/lukeocodes)
- [Bluesky](https://bsky.app/profile/lukeocodes.dev)
- [Email](mailto:luke@lukeoliff.com)
- [RSS](/rss.xml)

## Cookies

This site sets one analytics cookie, from Google Analytics, and only in a browser that accepts the cookie banner. A client that does not run scripts sets no cookies, so there is nothing to accept or decline here.

[Cookie policy](/cookies)
