Introducing Endeavor 1.1.
Intelligence to rival leading @OpenAI and @AnthropicAI models. Under your control.
A huge leap from Endeavor 1.0 to 1.1. Better reasoning, coding and agentic work. Multimodal is here. New EU GPU inference options with enterprise support.
Our third model release this year. Request access via link in thread.
Last day in Hong Kong for Flower’s Pedro Mesa.
Working on collaborative AI, privacy-preserving ML or frontier models on your own infrastructure? Talk to Pedro about @flwrlabs and Endeavor, frontier AI under your control.
DM us to connect.
1
110
How good is Endeavor for coding?
Endeavor scored a perfect 10/10 and finished second fastest in Paul Choi’s independent three-file coding test, matching @OpenAI’s GPT-6 Astra, @AnthropicAI’s Claude Opus 5.5 and @GoogleDeepMind’s Gemini 4 Argon on correctness.
The task involved fixing six related bugs. The 10 final tests were hidden from the models while coding, then run afterwards to check tricky cases and whether existing features still worked.
Paul tested Endeavor through the @flwrlabs API and shares the setup, results and execution logs.
The link to his full review is in the next post in this thread.
Want to try Endeavor on your own coding tasks? Message us.
1
109
Endeavor from @flwrlabs delivers a perfect 10/10 in Paul Choi’s three-file coding test, matching @OpenAI’s GPT-6 Astra and @AnthropicAI’s Claude Opus 5.5.
Read his independent review:
men16922.github.io/articles/…
55
Could the strongest AI models be trained across distributed data and compute?
Today, Flower’s @niclane shares early work from Flower Labs and @Cambridge_Uni’s @CaMLSys in his keynote at the 4th International Conference on Foundation and Large Language Models (FLLM 2026).
“Beyond the Mainframe Era of AI: Wide-Area Distributed Training at Frontier Scale”
Nic will make the case that federated and decentralized training will underpin the most capable future models.
He’ll explore the technical challenges of training across distributed data and compute, and early solutions that could help move AI beyond a centralized data-center model.
19 November · 13:00 CET · 12:00 GMT
Barcelona, Spain
If you’re attending FLLM 2026, catch Nic’s keynote.
Today, Flower’s Pedro Mesa joins the @UN panel “Developing open models and composing the stack” in Hong Kong.
The session is part of “The Economics of Open Development: The Case for Open Source AI,” a wider UN event Flower was invited to participate in alongside leading AI companies championing the domain.
@ODET_UN’s Mehdi Snene moderates. Pedro, our Founding Commercial Partner, joins @HongKongPolyU’s Prof. Hongxia Yang and Poman Lo, Founder of One Earth Initiatives, to explore how open AI can be built together.
For Flower, this connects directly to our work on collaborative AI, open-weight models and privacy-preserving ML. Our ambition is to advance collaborative superintelligence, connecting data and compute across organizations so more people can shape, own and benefit from AI.
19 November · 11:30–13:00 HKT · Hong Kong

ALT Flower at the United Nations event in Hong Kong. Developing open models and composing the stack, 19 November 2026, 11:30–13:00 HKT. Portraits of Pedro Mesa, Founding Commercial Partner at Flower Labs; Mehdi Snene, UN ODET moderator; Prof. Hongxia Yang of PolyU; and Poman Lo, Founder of One Earth Initiatives.
2
183
One hour to go at @Supercomputing #SC26.
Jerry Watkins from @SandiaLabs presents our 70B LLM federated training work, developed with Patrick Foley and the Flower team.
Today, Nov 18 · 11am Chicago time
@RedHat booth #2308, McCormick Place.
Today at @Supercomputing #SC26: Flower + @SandiaLabs on federated training of a 70B LLM across three US national labs.
Patrick Foley and the Flower team worked closely with Jerry Watkins and his colleagues at Sandia on training across Sandia, @Livermore_Lab and @LosAlamosNatLab.
NVIDIA + AMD clusters. Slurm + Flux.
Hear Jerry present our joint work:
Wed, Nov 18 · 11am Chicago time
@RedHat booth #2308, McCormick Place, Chicago.
Last day to meet Patrick Foley + Micah Sheller at the booth. Come say hello.
Jerry’s earlier Flower AI Day talk is in the reply below.
Watch Jerry Watkins’s earlier talk at Flower AI Day in San Francisco, with an introduction by Patrick Foley from Flower.
youtu.be/MeB89aH9K5E
1
56
Starting in 1 hour @Supercomputing: Patrick Foley's live demo of federated agents with @flwrlabs SuperGrid + @RedHat @openshift.
Tue, Nov 17 · 4pm Chicago time
Red Hat booth #2308, McCormick Place.
1
1
129
Today, Nov 17: meet Patrick Foley + Micah Sheller at @Supercomputing, @RedHat booth #2308
4pm Chicago time: Patrick's 15-minute demo of federated agents with @flwrlabs SuperGrid + @openshift
1
1
129
More from Patrick Foley -- his earlier talk, “Flower Confidential Compute,” recorded at Flower AI Summit in London. Highly related to his talk at @Supercomputing youtu.be/-kryTPxayhw
1
55
More from Micah Sheller: “Party Responsibly: Governing Cross-Platform Collaborative AI,” with Ann Kim Novakowski. Talk to Micah today @Supercomputing about governance. youtu.be/kKywME39U3E
47
Flower is one of a select few organizations invited to the @UN event on open-source AI in Hong Kong this Thursday, Nov 19: “The Economics of Open Development: The Case for Open Source AI”
Pedro Mesa, our Founding Commercial Partner, will join the panel on developing open models and composing the stack.
1
1
117
Flower + @SandiaLabs. A 70B LLM trained across three US national labs.
Patrick Foley and the Flower team worked closely with Jerry Watkins and colleagues at Sandia on federated training across Sandia, @Livermore_Lab and @LosAlamosNatLab.
NVIDIA + AMD clusters. Slurm + Flux.
Jerry presents our joint work at SC26
Wed, Nov 18 · 11am Chicago time
@RedHat booth #2308, McCormick Place, Chicago.
Come meet Patrick and the Flower team.
Tue, Nov 17, 4pm Chicago time @Supercomputing: Patrick Foley on federated agents with @flwrlabs SuperGrid + @RedHat @openshift. Booth #2308
A 15-minute talk + live demo using simulated telemetry: agents investigate together and share findings while source data stays under its owner's control.
1
1
50
Wed, Nov 18, 11am Chicago time @Supercomputing: @SandiaLabs's Jerry Watkins presents our federated training collaboration on a 70B LLM across Sandia, @Livermore_Lab and @LosAlamosNatLab.
NVIDIA + AMD clusters; Slurm + Flux.
Red Hat booth #2308, McCormick Place, Chicago. Come say hello.
50













