# mozilla-state-of-open-source-ai-2026-07

## Veille

**Recurring report from Mozilla**, *The state of open source AI*, **v1.0.1, July 2026**, introduced by a letter from **Raffi Krikorian** (CTO): seven sections, an interactive site, and a downloadable report. Thesis stated in the title of Section 1: *« The model layer has commoditized. Value accrues to the harness above it. »* **Capability state**: on the *Artificial Analysis Intelligence Index v4.1*, the best closed model scores **61** (Claude Opus 5) and the best open model **57** (**Kimi K3**), fourth overall and ahead of three of the largest closed labs; on the *Epoch Capabilities Index*, the gap is **6 points** (K3 at 156 versus GPT-5.6 Sol at 162), described as *« about one release cycle »*, with overlapping confidence intervals. **Sawtooth frontier**: open leads in frontend code (K3 at 1,679 Elo on LMArena Frontend Code Arena, six domains out of seven), contests agentic terminal work (88.3 versus 88.8 on Terminal-Bench 2.1), and cedes ground on professional knowledge work (Fable 5 leads K3 by 92 Elo on GDPval-AA v2). **Usage shift**: the share of OpenRouter tokens routed to open-weight models rose from a negligible level to a third by late 2025, then to a **majority by mid-2026**, with the seven highest-volume models all open-weight — the report itself noting that *« by request count, closed US providers still lead »*, the open lead being a token-volume lead concentrated in coding and agentic workloads. **The central contrast**: *« Open ships easy. Open deploys hard. »* — 79% of developers adding AI use open models versus 71% for closed, but only **53%** of open-model teams reach production **versus 63%**, and the gap widens with organization size (closed 54% → 73%, open 53% → 57%), which *« rules out a resources explanation »*. The stack maturity map (48 components, 9 layers) shows two consistently cold columns — **standardization** and ***enterprise readiness*** — identified as the operational gap. **Section 5**: *« The agentic harness is another user agent »*, and *« The model is eating the harness »* — on every model where both exist, the lab's own harness now wins, the 21.8-point gap having compressed to about 3. Hence the formula: *« A harness tuned tightly to one lab's weights… degrades on anyone else's model, so the tighter the tuning, the less swappable the weights underneath. Lock-in arrives as a side effect of optimization. »*

## Titre Article

The state of open source AI (v1.0.1, juillet 2026)

## Date

2026-07

## URL

https://stateofopensource.ai/

## Keywords

Mozilla, state of open source AI, open weights, open weights, open source AI, OSI definition, training code, data documentation, model commoditization, agentic harness, value upstream of the model, Artificial Analysis Intelligence Index, Epoch Capabilities Index, Kimi K3, Claude Opus 5, GPT-5.6 Sol, Fable 5, Terminal-Bench, GDPval, LMArena, sawtooth frontier, OpenRouter, token share, volume vs requests, inference pricing, 50x in 36 months, LLMflation, adoption vs production, Mozilla SlashData, operational gap, standardization, enterprise readiness, stack map, sovereignty, 70 national strategies, exit rights, exit rights, cloud exit cost, repatriation, Qwen, DeepSeek, Moonshot, Mistral, Zhipu, MiniMax, LangChain, MCP, A2A, Agentic AI Foundation, agent governance, unsolved write surface, lock-in through optimization, export controls, nineteen-day blackout, irreversibility of a weight release, distillation, Kratsios, Greenblatt, Redwood Research, watchlist, Raffi Krikorian

## Authors

**Mozilla** — éditeur du rapport, avec une introduction signée **Raffi Krikorian**, *Chief Technology Officer*. Publié en **juillet 2026** (v1.0.1). Données issues de sources tierces créditées (Artificial Analysis, Epoch AI, OpenRouter, LMArena) et d'une enquête propre menée avec **SlashData** (*Mozilla / SlashData 2026 developer survey*, n = 1 410 sur la question des freins).

Mozilla n'est pas un observateur neutre du sujet qu'elle mesure, et la lettre d'ouverture le revendique : *« Mozilla exists because one company tried to own the front door to the web, and an open community made sure it never could. We bet on open the first time. Open won. »* Le rapport est à la fois un travail de mesure sourcé et une pièce de plaidoyer, refermée sur un appel à l'action (*« Build with us »*) et une inscription à MozFest.

## Ton

**Profile**: **advocacy-driven sector state report**, interactive web format (clickable charts, hoverable cells) paired with a PDF. Analytical and data-heavy register in the body, **mobilizing at both ends** — the CTO's opening letter, the closing *« Build with us »* call.

**Style**: each section carries a **two-line thesis title** (*« Open ships easy. / Open deploys hard. »*, *« Open is a sovereignty choice. / Seventy governments are already treating it as one. »*), followed by a series of visualizations each captioned with its source **and its limitation**. This captioning discipline is what sets the document apart from a talking-points deck: *« Scores use different scales and come mostly from vendor-run tests, so treat them as directional »*, *« with intermediate points interpolated »*, *« the confidence intervals overlap »*, *« Shares are of the top 20 only, not of all OpenRouter traffic »*.

**Two structuring rhetorical moves**:

1. **Opening with the marginal case.** The letter opens on te reo māori — a language *« with no commercial market to speak of »* for which a New Zealand broadcaster is building speech models under a license that keeps recordings with the communities — and on a cassava-leaf diagnosis rendered by a phone-sized model in a connectivity-dead zone of East Africa. *« Neither project needed permission. And neither could have been rented from a frontier model. »* **The value of open is established through what the market will never fund, before market share is even mentioned.**
2. **The reversibility list.** The *watchlist* section does not merely list signals: each carries a ***« Reverses if: »*** — the condition that would invalidate the thesis. Rare and honest in an advocacy document.

**Marker phrases**: *« The model layer has commoditized »*, *« Open ships easy. Open deploys hard »*, *« Open weights are exit rights »*, *« You can switch off a model. You cannot switch off a copy already running on a machine you hold »*, *« the unsolved write surface »*, *« Lock-in arrives as a side effect of optimization »*, *« Commodity inputs surrender pricing power »*.

## Pense-betes

- **Date / source**: *The state of open source AI*, **Mozilla**, v1.0.1, **July 2026**, introduction by **Raffi Krikorian** (CTO). Third-party sources credited (Artificial Analysis, Epoch AI, OpenRouter, LMArena) and proprietary survey with **SlashData** (n = 1,410 on adoption barriers).
- **Key framing**: *« The model layer has commoditized. Value accrues to the harness above it. »* Justification: *« Commodity inputs surrender pricing power »*, and the majority of production workloads run well below the frontier ceiling. ### The terminology distinction, worth keeping | Term | What it covers | |---|---| | **Open model** | downloadable weights, executable and modifiable on hardware you control | | **Open weights** | parameters under a permissive license, **without training code or data documentation** — *« which describes most of what this report measures »* | | **Open source AI** (OSI sense) | additionally requires training code and enough data information to reconstruct the system | The report is titled *open source AI* but overwhelmingly measures **open weights**. It says so; the takeaways won't. ### Capability gap, two consistent instruments | Instrument | Closed | Open | Gap | |---|---|---|---| | Artificial Analysis Intelligence Index v4.1 | 61 (Claude Opus 5) | 57 (**Kimi K3**) | 4 pts; K3 4th out of 586 models, 3 of the top 11 open-weight | | Epoch Capabilities Index | 162 (GPT-5.6 Sol) | 156 (K3) | 6 pts ≈ one release cycle, **overlapping confidence intervals** | On price, K3 sits 3.6 points off the top for roughly a third of the price. ### The sawtooth frontier Open **leads** in frontend code (K3, 1,679 Elo, six domains out of seven on LMArena Frontend Code Arena); **contests** terminal agentic work (88.3 versus 88.8 on Terminal-Bench 2.1; K3 wins Program Bench, SpreadsheetBench 2, and BrowseComp, loses FrontierSWE 81.2 versus 86.6); **cedes** ground on professional knowledge work (Fable 5 leads by **92 Elo** on GDPval-AA v2, the largest gap among shared benchmarks, and Moonshot concedes it). Hence: *« Match the model to the job and you need the frontier for less than you think. »* The report warns that scores use different scales and mostly come from vendor-run tests — *« treat them as directional »*. ### The most misquoted figure Open weights rose from a negligible base to **a third** of OpenRouter tokens by late 2025, then to a **majority** by mid-2026, with the seven highest-volume models all open-weight (72.4% of top-20 volume held by open-weight ranks 1-10). But: *« By request count, closed US providers still lead. The open lead is a token-volume lead, concentrated in coding and agentic workloads »*, and the shares cover only the **top 20**. A majority of tokens is not a majority of use cases. ### Deployment over capability *« Open ships easy. Open deploys hard. »* 79% of developers adding AI use open models versus 71% closed, and **50% use both** (29% open only, 21% closed only): the two categories are complementary, not rival. But 53% of open teams reach production versus 63% closed, and the gap widens with size — closed **54% → 73%**, open **53% → 57%**. *« Scale rules out a resources explanation. Enterprises can buy their way through closed deployment. Open deployment waits on tooling that remains unfinished. »* Named barriers, all operational: infrastructure cost, security and compliance, maintenance, deployment complexity. The stack map (48 components, 9 layers, 10 criteria) confirms this from another angle: two consistently cold columns across every layer — **standardization** and ***enterprise readiness***. ### Section 5: the harness as user agent The analogy is the browser, *« code on the user's side negotiating with servers on their behalf »*, replayed one layer up. The harness is *« where production difficulty concentrates, and where the open-vs-closed, owner-vs-renter contest restarts »*. Five mapped layers: **Govern** (stateful policy, registry and lineage, budget and revocation), **Surface** (AG-UI/A2UI, x402/AP2/UCP), **Action** (E2B/Daytona/Modal sandboxes, permission and identity — *« the unsolved write surface »* —, eval and observability), **Reach** (MCP, A2A, memory), **Control** (LangGraph, CrewAI, AutoGen, LlamaIndex). The phrase *« the unsolved write surface »* echoes the diagnosis in [[valente-zalewski-beyond-zero-enterprise-security-ai-era-2026-07-20]]: read is solved, write is not. ### The lock-in mechanism *« The model is eating the harness. »* On every model where a lab's own harness and an independent harness coexist, the former now wins, the 21.8-point gap having compressed to about 3 at the top. Hence: *« A harness tuned tightly to one lab's weights becomes a fitted component of that lab's product. It degrades on anyone else's model, so the tighter the tuning, the less swappable the weights underneath. Lock-in arrives as a side effect of optimization. »* Lock-in doesn't need to be a strategy: optimizing is enough. Compare with the portability contract in [[janakiram-agent-platform-portability-contract-2026-07-20]]. ### Traction and economics LangChain 126,000+ stars and 60% developer share; **MCP** at 97M monthly SDK downloads and 10,000+ active servers within a year, **+4,750% in 16 months**, handed to the Agentic AI Foundation in December 2025. Governance gap: **only ~21%** of enterprises report mature agent governance. On pricing, inference dropped **50x in 36 months** at GPT-4 level (versus 2.6x for bandwidth during the dotcom era), the frontier price having fallen 112x since GPT-4's $45. On OpenRouter (May-Sept. 2025), closed held ~80% of usage and ~96% of revenue — at quality parity, it costs roughly **6x more per call**, implying an estimated **~$24.8B in unrealized annual savings** (Nagle-Yue study for the Linux Foundation). ### Sovereignty and the right to exit More than **70 active national strategies**, and *« the strategic case for open is the ability to leave »*, backed by the cloud precedent ($90-120K to exit a petabyte from S3, 37signals down from $3.2M to under $1M, 80% of enterprises repatriating). *« Closed model APIs reproduce the same trap… Open weights are exit rights. »* **The nineteen-day blackout** makes the argument concrete: June 9, Anthropic ships Fable 5 and Mythos 5 → **June 12**, Commerce applies export controls with immediate effect, barring access to any foreign national, including Anthropic's own employees; nationality being unverifiable in real time, **both models go dark for everyone** → June 26, partial restoration of Mythos for ~100 US critical-infrastructure organizations → June 30, lifted → July 1, Fable 5 restored. Then July 16, Moonshot opens K3's API. *« Access can be revoked and restored. A weight release cannot be withdrawn once the files are distributed… You can switch off a model. You cannot switch off a copy already running on a machine you hold. »* ### China Qwen surpassed the next eight organizations combined in Hugging Face downloads by February 2026; Chinese open-weight models rose from under 2% of OpenRouter tokens in late 2024 to over **45%** of weekly traffic by April 2026, ~61% among the ten most-used models. DeepSeek claims 26,000+ enterprise accounts and was part of 58% of new AI startups' stacks in 2025, even as at least eight jurisdictions restricted the hosted service: *« Enterprises ban the hosted app and adopt the weights anyway. »* ### The K3 distillation affair, by the report's own categories | Level | Content | |---|---| | **Confirmed** | Anthropic's February 2026 disclosure — ~24,000 fraudulent accounts, over 16M exchanges of which **3.4M attributed to Moonshot**, against earlier Claude models; statements from Kratsios (July 22) and Bessent (21) | | **Signal** | Greenblatt (Redwood Research): K3 self-identifies as Claude in a way statistically hard to explain as noise — but it names a model predating the affair, and self-identification is a known artifact of training on web text | | **Proof** | **absent** — *« No logs and no forensic package. Moonshot denies. »* | The report notes that weight publication now makes behavioral forensics possible. The allegation should never be presented as established. ### The watchlist and its method Four families of signals (capability/adoption, harness, market structure, trust/safety), each paired with its own **reversal condition** — *« Reverses if: the lab-harness lead widens, or a closed platform sets the permission standard first »*. Stating in advance what would prove the thesis wrong is what sets this document apart from an ordinary advocacy piece. ### A note on citation caution Mozilla is measuring a subject it advocates for. The rigor of its captions and the presence of reversal conditions make the data usable; the framing — commoditization achieved, openness inevitable, *« open won »* — is a thesis, not a finding.

## RésuméDe400mots

Recurring report from **Mozilla**, *The state of open source AI* (v1.0.1, July 2026), introduced by its CTO **Raffi Krikorian**.

**The thesis** opens the first section: *« The model layer has commoditized. Value accrues to the harness above it. »* Inputs that have become commodities lose their pricing power, and the majority of production workloads run well below the frontier ceiling.

**Capability state.** On the Artificial Analysis Intelligence Index, the best closed model scores 61 (Claude Opus 5), the best open model 57 (**Kimi K3**), fourth overall; on the Epoch Capabilities Index the gap is **six points, "about one release cycle"**, with overlapping confidence intervals. The frontier is **sawtooth**: open leads in frontend code, contests terminal agentic work, and clearly cedes ground on professional knowledge work.

**The usage shift.** The share of OpenRouter tokens routed to open weights rose from a negligible level to a majority by mid-2026, with the seven highest-volume models all open — but the report notes that **by request count, closed providers still lead**, the open lead being a token-volume lead concentrated in coding and agentic workloads.

**The central finding**: *« Open ships easy. Open deploys hard. »* 79% of developers use open models versus 71% closed, with half using both; but only **53% of open teams reach production versus 63%**, and the gap **widens with company size**, which rules out an explanation by resources. The stack map confirms it: two cold columns across every layer, **standardization and *enterprise readiness***.

**The harness is the new frontier.** *« The agentic harness is another user agent »* — the browser's role replayed one layer up. And the lock-in mechanism is stated precisely: a lab's harness, tuned to its own weights, degrades on everyone else's, so *« the tighter the tuning, the less swappable the weights underneath. **Lock-in arrives as a side effect of optimization.** »*

**Sovereignty** is framed as a right to exit, illustrated by Fable 5's **nineteen-day blackout** over export controls: *« You can switch off a model. You cannot switch off a copy already running on a machine you hold. »*

Mozilla advocates for what it measures. Scrupulous captions and a self-stated reversal watchlist make the data usable; the framing remains a thesis.

## GrapheDeConnaissance

- Mozilla —publie→ The state of open source AI (DOCUMENT, 0.98)
- Raffi Krikorian —travaille_chez→ Mozilla (ORGANISATION, 0.95)
- Mozilla —affirme_que→ la couche modèle s'est commoditisée et la valeur remonte vers le harnais agentique (CITATION, 0.96)
- harnais agentique —est_instance_de→ un user agent, au sens où le navigateur l'était pour le web ouvert (AFFIRMATION, 0.92)
- harnais ajusté aux poids d'un laboratoire —réduit→ l'interchangeabilité des poids sous-jacents — le verrouillage arrive comme effet de bord de l'optimisation (CITATION, 0.95)
- Kimi K3 —mesure→ 57 sur l'Artificial Analysis Intelligence Index v4.1, contre 61 pour le meilleur modèle fermé (MESURE, 0.93)
- Kimi K3 —concurrence→ Claude Opus 5 (TECHNOLOGIE, 0.9)
- open-weights —mesure→ une majorité des tokens routés sur OpenRouter mi-2026, avec les sept plus forts volumes (MESURE, 0.9)
- Mozilla —affirme_que→ l'avance des poids ouverts est un volume de tokens et non un nombre de requêtes, concentré sur le codage et l'agentique (AFFIRMATION, 0.93)
- modèles ouverts —s_oppose_à→ leur propre mise en production, seules 53 % des équipes y parvenant contre 63 % en fermé (MESURE, 0.92)
- écart opérationnel —observé_dans→ la standardisation et l'enterprise readiness, colonnes systématiquement faibles de toutes les couches du stack ouvert (AFFIRMATION, 0.93)
- poids ouverts —permet→ un droit de sortie, là où une API propriétaire reproduit le piège du verrouillage cloud (CITATION, 0.94)
- contrôle à l'export —s_oppose_à→ la disponibilité d'un modèle fermé, éteint dix-neuf jours pour tous ses utilisateurs (AFFIRMATION, 0.94)
- publication de poids —s_oppose_à→ la révocabilité d'un accès, une copie distribuée ne pouvant être retirée (CITATION, 0.95)
- permission et identité des agents —est_instance_de→ la surface d'écriture non résolue du stack agentique (CITATION, 0.93)
- Model Context Protocol —mesure→ 97 M de téléchargements SDK mensuels et plus de 10 000 serveurs actifs en un an (MESURE, 0.9)
- gouvernance des agents —s_oppose_à→ le rythme d'adoption, seules ~21 % des entreprises déclarant une gouvernance mature (MESURE, 0.9)
- prix de l'inférence —réduit→ d'un facteur 50 en 36 mois à performance de classe GPT-4 (MESURE, 0.9)
- modèles fermés —mesure→ environ 80 % de l'usage et 96 % du revenu sur OpenRouter, à un coût par appel environ six fois supérieur à parité (MESURE, 0.88)
- entreprises —utilise→ les poids ouverts chinois en auto-hébergement, tout en interdisant le service hébergé correspondant (AFFIRMATION, 0.88)
- Michael Kratsios —affirme_que→ Kimi K3 aurait été entraîné par extraction covert à grande échelle depuis Fable 5 (AFFIRMATION, 0.85)
- Moonshot AI —s_oppose_à→ l'allégation d'extraction covert, qu'elle dément (AFFIRMATION, 0.9)
- Mozilla —affirme_que→ aucun journal ni dossier forensique n'étaye l'allégation de distillation, que Moonshot dément (AFFIRMATION, 0.92)
- Mozilla —recommande→ de suivre quatre familles de signaux assorties chacune de sa condition d'invalidation (AFFIRMATION, 0.9)
- open-weights —s_oppose_à→ open source AI au sens OSI, qui exige en plus le code d'entraînement et la documentation des données (AFFIRMATION, 0.95)

---
Canonical: https://www.thekb.eu/en/fiches/mozilla-state-of-open-source-ai-2026-07/