#apertus — Public Fediverse posts

@vincent @wbpeckham @cxiao I struggle to find any ethical flaws with #Apertus at the moment.

Human marking of abusive content was a nasty response to demands to impose censorship (euphemistically called "guardrails") into #AI .

@vincent @wbpeckham @cxiao I struggle to find any ethical flaws with #Apertus at the moment.

Human marking of abusive content was a nasty response to demands to impose censorship (euphemistically called "guardrails") into #AI .

@vincent @wbpeckham @cxiao I struggle to find any ethical flaws with #Apertus at the moment.

Human marking of abusive content was a nasty response to demands to impose censorship (euphemistically called "guardrails") into #AI .

@vincent @wbpeckham @cxiao I struggle to find any ethical flaws with #Apertus at the moment.

Human marking of abusive content was a nasty response to demands to impose censorship (euphemistically called "guardrails") into #AI .

Familienforschung in Rheinhessen

https://www.compgen.de/2026/08/familienforschung-in-rheinhessen/

Familienforschung in Rheinhessen

https://www.compgen.de/2026/08/familienforschung-in-rheinhessen/

Familienforschung in Rheinhessen

https://www.compgen.de/2026/08/familienforschung-in-rheinhessen/

Thanks to the 50+ contributors helping to launch Public AI Switzerland at the grassroots - a cooperative that embodies open standards, respects user data sovereignty and embodies shared economy principles - aiming to serve the common interest as a trusted public utility. 

Initiated with the support of the international #PublicAI Network, based on the #Apertus model, aligned to many of the same principles that make ActivityPub and the Open Web work for millions of people today.

#participer #fatschpart #partecipare #mitmachen #joinin https://publicai.ch

Thanks to the 50+ contributors helping to launch Public AI Switzerland at the grassroots - a cooperative that embodies open standards, respects user data sovereignty and embodies shared economy principles - aiming to serve the common interest as a trusted public utility. 

Initiated with the support of the international #PublicAI Network, based on the #Apertus model, aligned to many of the same principles that make ActivityPub and the Open Web work for millions of people today.

#participer #fatschpart #partecipare #mitmachen #joinin https://publicai.ch

Thanks to the 50+ contributors helping to launch Public AI Switzerland at the grassroots - a cooperative that embodies open standards, respects user data sovereignty and embodies shared economy principles - aiming to serve the common interest as a trusted public utility. 

Initiated with the support of the international #PublicAI Network, based on the #Apertus model, aligned to many of the same principles that make ActivityPub and the Open Web work for millions of people today.

#participer #fatschpart #partecipare #mitmachen #joinin https://publicai.ch

Thanks to the 50+ contributors helping to launch Public AI Switzerland at the grassroots - a cooperative that embodies open standards, respects user data sovereignty and embodies shared economy principles - aiming to serve the common interest as a trusted public utility. 

Initiated with the support of the international #PublicAI Network, based on the #Apertus model, aligned to many of the same principles that make ActivityPub and the Open Web work for millions of people today.

#participer #fatschpart #partecipare #mitmachen #joinin https://publicai.ch

Thanks to the 50+ contributors helping to launch Public AI Switzerland at the grassroots - a cooperative that embodies open standards, respects user data sovereignty and embodies shared economy principles - aiming to serve the common interest as a trusted public utility. 

Initiated with the support of the international #PublicAI Network, based on the #Apertus model, aligned to many of the same principles that make ActivityPub and the Open Web work for millions of people today.

#participer #fatschpart #partecipare #mitmachen #joinin https://publicai.ch

Moving toward open-weight AI offers both accountability and efficiency

While mega #datacenters dominate, efficiency gains remain elusive. Decentralization may indeed address AI's mounting energy issues and addressing specialized demands in #scientific applications.

Alternative (more) #sustainable #llm: fully open, transparent, locally-deployable, and compliant with data protection and copyrights, as exemplified by #apertus by Swiss institutions

https://www.nature.com/articles/d41586-026-02451-2

@apertus #leadership

Moving toward open-weight AI offers both accountability and efficiency

While mega #datacenters dominate, efficiency gains remain elusive. Decentralization may indeed address AI's mounting energy issues and addressing specialized demands in #scientific applications.

Alternative (more) #sustainable #llm: fully open, transparent, locally-deployable, and compliant with data protection and copyrights, as exemplified by #apertus by Swiss institutions

https://www.nature.com/articles/d41586-026-02451-2

@apertus #leadership

Moving toward open-weight AI offers both accountability and efficiency

While mega #datacenters dominate, efficiency gains remain elusive. Decentralization may indeed address AI's mounting energy issues and addressing specialized demands in #scientific applications.

Alternative (more) #sustainable #llm: fully open, transparent, locally-deployable, and compliant with data protection and copyrights, as exemplified by #apertus by Swiss institutions

https://www.nature.com/articles/d41586-026-02451-2

@apertus #leadership

Moving toward open-weight AI offers both accountability and efficiency

While mega #datacenters dominate, efficiency gains remain elusive. Decentralization may indeed address AI's mounting energy issues and addressing specialized demands in #scientific applications.

Alternative (more) #sustainable #llm: fully open, transparent, locally-deployable, and compliant with data protection and copyrights, as exemplified by #apertus by Swiss institutions

https://www.nature.com/articles/d41586-026-02451-2

@apertus #leadership

Moving toward open-weight AI offers both accountability and efficiency

While mega #datacenters dominate, efficiency gains remain elusive. Decentralization may indeed address AI's mounting energy issues and addressing specialized demands in #scientific applications.

Alternative (more) #sustainable #llm: fully open, transparent, locally-deployable, and compliant with data protection and copyrights, as exemplified by #apertus by Swiss institutions

https://www.nature.com/articles/d41586-026-02451-2

@apertus #leadership

Schweizer Daten an Schweizer Modelle 👍

#Apertus
(Der Rest nicht so toll.)
https://www.tagesanzeiger.ch/srg-blockiert-ki-bots-archiv-nur-noch-fuer-apertus-679758954167

Schweizer Daten an Schweizer Modelle 👍

#Apertus
(Der Rest nicht so toll.)
https://www.tagesanzeiger.ch/srg-blockiert-ki-bots-archiv-nur-noch-fuer-apertus-679758954167

Schweizer Daten an Schweizer Modelle 👍

#Apertus
(Der Rest nicht so toll.)
https://www.tagesanzeiger.ch/srg-blockiert-ki-bots-archiv-nur-noch-fuer-apertus-679758954167

Schweizer Daten an Schweizer Modelle 👍

#Apertus
(Der Rest nicht so toll.)
https://www.tagesanzeiger.ch/srg-blockiert-ki-bots-archiv-nur-noch-fuer-apertus-679758954167

Schweizer Daten an Schweizer Modelle 👍

#Apertus
(Der Rest nicht so toll.)
https://www.tagesanzeiger.ch/srg-blockiert-ki-bots-archiv-nur-noch-fuer-apertus-679758954167

I think the fine tuning, and model specialization here is really what is interesting.

But this should be achievable with open models that are more ethical like #swissai #apertus 1.5

https://www.youtube.com/watch?v=-IGB6Avxwgo

#ai

I think the fine tuning, and model specialization here is really what is interesting.

But this should be achievable with open models that are more ethical like #swissai #apertus 1.5

https://www.youtube.com/watch?v=-IGB6Avxwgo

#ai

I think the fine tuning, and model specialization here is really what is interesting.

But this should be achievable with open models that are more ethical like #swissai #apertus 1.5

https://www.youtube.com/watch?v=-IGB6Avxwgo

#ai

I think the fine tuning, and model specialization here is really what is interesting.

But this should be achievable with open models that are more ethical like #swissai #apertus 1.5

https://www.youtube.com/watch?v=-IGB6Avxwgo

#ai

I think the fine tuning, and model specialization here is really what is interesting.

But this should be achievable with open models that are more ethical like #swissai #apertus 1.5

https://www.youtube.com/watch?v=-IGB6Avxwgo

#ai

SRG-Chefin #SusanneWille im #NZZS-Interview:

Seit August weist die SRG KI-#Trainingsbots ab, künftig soll ihr #Archiv dem Schweizer #KIModell #Apertus dienen.

Ihr Fazit: „Unser Journalismus ist etwas wert."

Als Vorbild dient Frankreichs „Must-be-found"-Regel für öffentlich-rechtliche Inhalte.

Und eine Mahnung mit Verweis auf den Club of Rome: Die grösste Gefahr sei nicht Krieg, sondern die Unfähigkeit, #Fakten von #Fiktion zu unterscheiden.

#SRG #KI #Medien

https://www.linkedin.com/posts/activity-7492223037610725378-jhrW?utm_source=share&utm_medium=member_desktop&rcm=ACoAABIszQcBLNYFSd2mvhf5aR9Z50UI95I6Oow

SRG-Chefin #SusanneWille im #NZZS-Interview:

Seit August weist die SRG KI-#Trainingsbots ab, künftig soll ihr #Archiv dem Schweizer #KIModell #Apertus dienen.

Ihr Fazit: „Unser Journalismus ist etwas wert."

Als Vorbild dient Frankreichs „Must-be-found"-Regel für öffentlich-rechtliche Inhalte.

Und eine Mahnung mit Verweis auf den Club of Rome: Die grösste Gefahr sei nicht Krieg, sondern die Unfähigkeit, #Fakten von #Fiktion zu unterscheiden.

#SRG #KI #Medien

https://www.linkedin.com/posts/activity-7492223037610725378-jhrW?utm_source=share&utm_medium=member_desktop&rcm=ACoAABIszQcBLNYFSd2mvhf5aR9Z50UI95I6Oow

SRG-Chefin #SusanneWille im #NZZS-Interview:

Seit August weist die SRG KI-#Trainingsbots ab, künftig soll ihr #Archiv dem Schweizer #KIModell #Apertus dienen.

Ihr Fazit: „Unser Journalismus ist etwas wert."

Als Vorbild dient Frankreichs „Must-be-found"-Regel für öffentlich-rechtliche Inhalte.

Und eine Mahnung mit Verweis auf den Club of Rome: Die grösste Gefahr sei nicht Krieg, sondern die Unfähigkeit, #Fakten von #Fiktion zu unterscheiden.

#SRG #KI #Medien

https://www.linkedin.com/posts/activity-7492223037610725378-jhrW?utm_source=share&utm_medium=member_desktop&rcm=ACoAABIszQcBLNYFSd2mvhf5aR9Z50UI95I6Oow

SRG-Chefin #SusanneWille im #NZZS-Interview:

Seit August weist die SRG KI-#Trainingsbots ab, künftig soll ihr #Archiv dem Schweizer #KIModell #Apertus dienen.

Ihr Fazit: „Unser Journalismus ist etwas wert."

Als Vorbild dient Frankreichs „Must-be-found"-Regel für öffentlich-rechtliche Inhalte.

Und eine Mahnung mit Verweis auf den Club of Rome: Die grösste Gefahr sei nicht Krieg, sondern die Unfähigkeit, #Fakten von #Fiktion zu unterscheiden.

#SRG #KI #Medien

https://www.linkedin.com/posts/activity-7492223037610725378-jhrW?utm_source=share&utm_medium=member_desktop&rcm=ACoAABIszQcBLNYFSd2mvhf5aR9Z50UI95I6Oow

SRG-Chefin #SusanneWille im #NZZS-Interview:

Seit August weist die SRG KI-#Trainingsbots ab, künftig soll ihr #Archiv dem Schweizer #KIModell #Apertus dienen.

Ihr Fazit: „Unser Journalismus ist etwas wert."

Als Vorbild dient Frankreichs „Must-be-found"-Regel für öffentlich-rechtliche Inhalte.

Und eine Mahnung mit Verweis auf den Club of Rome: Die grösste Gefahr sei nicht Krieg, sondern die Unfähigkeit, #Fakten von #Fiktion zu unterscheiden.

#SRG #KI #Medien

https://www.linkedin.com/posts/activity-7492223037610725378-jhrW?utm_source=share&utm_medium=member_desktop&rcm=ACoAABIszQcBLNYFSd2mvhf5aR9Z50UI95I6Oow

@geerlingguy I am using a #pi4 with #Mistral 3B as my document classifier. I would prefer #Apertus but while 8B outperforms Mistral 3B, its too heavy for the Pi4 and the 1.5B/4B variations quality is not sufficient https://codeberg.org/lsmith/nextcloud-classification/src/branch/experiment/eval-force-json/docs/apertus-mini-evaluation.md

So I am intrigued by testing out a #pi5 16GB with Apertus 8B. FTR Apertus v1.5 8B isn't yet ready for a Pi5 deployment https://mastodon.green/@lsmith/117054219745284915 //cc @loleg

@geerlingguy I am using a #pi4 with #Mistral 3B as my document classifier. I would prefer #Apertus but while 8B outperforms Mistral 3B, its too heavy for the Pi4 and the 1.5B/4B variations quality is not sufficient https://codeberg.org/lsmith/nextcloud-classification/src/branch/experiment/eval-force-json/docs/apertus-mini-evaluation.md

So I am intrigued by testing out a #pi5 16GB with Apertus 8B. FTR Apertus v1.5 8B isn't yet ready for a Pi5 deployment https://mastodon.green/@lsmith/117054219745284915 //cc @loleg

@geerlingguy I am using a #pi4 with #Mistral 3B as my document classifier. I would prefer #Apertus but while 8B outperforms Mistral 3B, its too heavy for the Pi4 and the 1.5B/4B variations quality is not sufficient https://codeberg.org/lsmith/nextcloud-classification/src/branch/experiment/eval-force-json/docs/apertus-mini-evaluation.md

So I am intrigued by testing out a #pi5 16GB with Apertus 8B. FTR Apertus v1.5 8B isn't yet ready for a Pi5 deployment https://mastodon.green/@lsmith/117054219745284915 //cc @loleg

@geerlingguy I am using a #pi4 with #Mistral 3B as my document classifier. I would prefer #Apertus but while 8B outperforms Mistral 3B, its too heavy for the Pi4 and the 1.5B/4B variations quality is not sufficient https://codeberg.org/lsmith/nextcloud-classification/src/branch/experiment/eval-force-json/docs/apertus-mini-evaluation.md

So I am intrigued by testing out a #pi5 16GB with Apertus 8B. FTR Apertus v1.5 8B isn't yet ready for a Pi5 deployment https://mastodon.green/@lsmith/117054219745284915 //cc @loleg

@geerlingguy I am using a #pi4 with #Mistral 3B as my document classifier. I would prefer #Apertus but while 8B outperforms Mistral 3B, its too heavy for the Pi4 and the 1.5B/4B variations quality is not sufficient https://codeberg.org/lsmith/nextcloud-classification/src/branch/experiment/eval-force-json/docs/apertus-mini-evaluation.md

So I am intrigued by testing out a #pi5 16GB with Apertus 8B. FTR Apertus v1.5 8B isn't yet ready for a Pi5 deployment https://mastodon.green/@lsmith/117054219745284915 //cc @loleg

Here's a capture from the one-shot prompt using #OpenCode to "Develop a small animation in a single HTML file with a vanilla JavaScript block. It should create an animated Matrix-like effect on the screen." #Apertus

Here's a capture from the one-shot prompt using #OpenCode to "Develop a small animation in a single HTML file with a vanilla JavaScript block. It should create an animated Matrix-like effect on the screen." #Apertus

Here's a capture from the one-shot prompt using #OpenCode to "Develop a small animation in a single HTML file with a vanilla JavaScript block. It should create an animated Matrix-like effect on the screen." #Apertus

Here's a capture from the one-shot prompt using #OpenCode to "Develop a small animation in a single HTML file with a vanilla JavaScript block. It should create an animated Matrix-like effect on the screen." #Apertus

Here's a capture from the one-shot prompt using #OpenCode to "Develop a small animation in a single HTML file with a vanilla JavaScript block. It should create an animated Matrix-like effect on the screen." #Apertus

Quick vibe check of #Apertus 1.5 70B running on the #OpenHands community-maintained TUI,. Suggestions of other tools, benchmarks and policies welcome

Quick vibe check of #Apertus 1.5 70B running on the #OpenHands community-maintained TUI,. Suggestions of other tools, benchmarks and policies welcome

Quick vibe check of #Apertus 1.5 70B running on the #OpenHands community-maintained TUI,. Suggestions of other tools, benchmarks and policies welcome

Quick vibe check of #Apertus 1.5 70B running on the #OpenHands community-maintained TUI,. Suggestions of other tools, benchmarks and policies welcome

Quick vibe check of #Apertus 1.5 70B running on the #OpenHands community-maintained TUI,. Suggestions of other tools, benchmarks and policies welcome

Anastasia Stasenko hits the nail on the head, addressing fairness and equity issues with the economic models of AI projects, not least the leadership opportunities for those implementing #Apertus or her own #Pleias LLM #POSAIS26

Anastasia Stasenko hits the nail on the head, addressing fairness and equity issues with the economic models of AI projects, not least the leadership opportunities for those implementing #Apertus or her own #Pleias LLM #POSAIS26

Anastasia Stasenko hits the nail on the head, addressing fairness and equity issues with the economic models of AI projects, not least the leadership opportunities for those implementing #Apertus or her own #Pleias LLM #POSAIS26

Anastasia Stasenko hits the nail on the head, addressing fairness and equity issues with the economic models of AI projects, not least the leadership opportunities for those implementing #Apertus or her own #Pleias LLM #POSAIS26

Anastasia Stasenko hits the nail on the head, addressing fairness and equity issues with the economic models of AI projects, not least the leadership opportunities for those implementing #Apertus or her own #Pleias LLM #POSAIS26

Open AI model #Apertus performs rather well in the (non-scientific) "city guide" benchmark:

"The question is not "sovereign AI or quality" but "how much quality gap are you willing to accept for full sovereignty?" And that gap is closing."

https://www.liip.ch/de/blog/apertus-after-8-months-what-we-learned-and-look-forward-using-switzerlands-ai-model

Open AI model #Apertus performs rather well in the (non-scientific) "city guide" benchmark:

"The question is not "sovereign AI or quality" but "how much quality gap are you willing to accept for full sovereignty?" And that gap is closing."

https://www.liip.ch/de/blog/apertus-after-8-months-what-we-learned-and-look-forward-using-switzerlands-ai-model

Open AI model #Apertus performs rather well in the (non-scientific) "city guide" benchmark:

"The question is not "sovereign AI or quality" but "how much quality gap are you willing to accept for full sovereignty?" And that gap is closing."

https://www.liip.ch/de/blog/apertus-after-8-months-what-we-learned-and-look-forward-using-switzerlands-ai-model

Open AI model #Apertus performs rather well in the (non-scientific) "city guide" benchmark:

"The question is not "sovereign AI or quality" but "how much quality gap are you willing to accept for full sovereignty?" And that gap is closing."

https://www.liip.ch/de/blog/apertus-after-8-months-what-we-learned-and-look-forward-using-switzerlands-ai-model

Open AI model #Apertus performs rather well in the (non-scientific) "city guide" benchmark:

"The question is not "sovereign AI or quality" but "how much quality gap are you willing to accept for full sovereignty?" And that gap is closing."

https://www.liip.ch/de/blog/apertus-after-8-months-what-we-learned-and-look-forward-using-switzerlands-ai-model

Response of Apertus 1.1 4B, on a RAG that includes documentation of the #Apertus project, and the recently published preliminary report from the UN Scientific Panel on AI:

~ Actionable Recommendation ~

Encourage the Swiss AI Initiative to develop a National Action Plan that:

A. Requires pilot projects with clear, measurable transparency targets linked to local authorities and public engagement.

B. Mandates the development of implementation guidelines with case studies that assess compliance, including provisions for data quality and global interoperability.

C. Establishes an annual reporting mechanism, not just to Swiss authorities but to the public and local governments outside of Switzerland.

After such a plan is implemented, the Swiss AI Initiative should conduct an inter-cantonal review assessing alignment across all cantons.

Civil Society Organizations engaged in research, community projects, or who experience AI adoption should be brought into the process to feedback directly on:

- The feasibility of requirements (e.g., data volume, technical complexity) in their respective contexts.
- The balance between compliance and innovation (potential for slowing the pace of adoption).

The Swiss AI Charter provides a solid theoretical foundation, but its operational applicability requires nuanced implementation that takes into account varying national legal and socio-cultural frameworks.

Actioning this will help the initiative build pragmatic frameworks that balance universality with adaptability, avoiding overregulation in some contexts while ensuring core principles are not eroded.

Response of Apertus 1.1 4B, on a RAG that includes documentation of the #Apertus project, and the recently published preliminary report from the UN Scientific Panel on AI:

~ Actionable Recommendation ~

Encourage the Swiss AI Initiative to develop a National Action Plan that:

A. Requires pilot projects with clear, measurable transparency targets linked to local authorities and public engagement.

B. Mandates the development of implementation guidelines with case studies that assess compliance, including provisions for data quality and global interoperability.

C. Establishes an annual reporting mechanism, not just to Swiss authorities but to the public and local governments outside of Switzerland.

After such a plan is implemented, the Swiss AI Initiative should conduct an inter-cantonal review assessing alignment across all cantons.

Civil Society Organizations engaged in research, community projects, or who experience AI adoption should be brought into the process to feedback directly on:

- The feasibility of requirements (e.g., data volume, technical complexity) in their respective contexts.
- The balance between compliance and innovation (potential for slowing the pace of adoption).

The Swiss AI Charter provides a solid theoretical foundation, but its operational applicability requires nuanced implementation that takes into account varying national legal and socio-cultural frameworks.

Actioning this will help the initiative build pragmatic frameworks that balance universality with adaptability, avoiding overregulation in some contexts while ensuring core principles are not eroded.

Response of Apertus 1.1 4B, on a RAG that includes documentation of the #Apertus project, and the recently published preliminary report from the UN Scientific Panel on AI:

~ Actionable Recommendation ~

Encourage the Swiss AI Initiative to develop a National Action Plan that:

A. Requires pilot projects with clear, measurable transparency targets linked to local authorities and public engagement.

B. Mandates the development of implementation guidelines with case studies that assess compliance, including provisions for data quality and global interoperability.

C. Establishes an annual reporting mechanism, not just to Swiss authorities but to the public and local governments outside of Switzerland.

After such a plan is implemented, the Swiss AI Initiative should conduct an inter-cantonal review assessing alignment across all cantons.

Civil Society Organizations engaged in research, community projects, or who experience AI adoption should be brought into the process to feedback directly on:

- The feasibility of requirements (e.g., data volume, technical complexity) in their respective contexts.
- The balance between compliance and innovation (potential for slowing the pace of adoption).

The Swiss AI Charter provides a solid theoretical foundation, but its operational applicability requires nuanced implementation that takes into account varying national legal and socio-cultural frameworks.

Actioning this will help the initiative build pragmatic frameworks that balance universality with adaptability, avoiding overregulation in some contexts while ensuring core principles are not eroded.