#openweights
Neither Deepseek or Metas models are opensource. They are openweights. Don’t let them redefine what open source means.
Friends, for something to be open source, we need to see

1. The data it was trained & evaluated on

2. The code

3. Model architecture

4. Model weights.

DeepSeek only gives 3, 4. And I'll see the day that anyone gives us #1 without being forced to do so, because all of them are stealing data.
January 29, 2025 at 8:34 PM
OpenWeights und lokale Inferenz kommen im Paper nicht vor. Das ist keine Lücke. Ein Verkäufer erwähnt das Produkt nicht, das seinen Markt zerstört.
June 29, 2026 at 9:26 AM
suppose i can download an openweights, open training set, 320B parameter model. i can't actually run it
December 7, 2025 at 6:48 PM
he's mad about openweights because well, they're ramming his employer pretty badly and are "irresponsible"
September 25, 2026 at 11:49 PM
Europa hat Energie. Europa hat Hardware — jedes moderne Endgerät rechnet Modelle lokal. OpenWeights-Modelle laufen heute darauf, ohne Rechenzentrum, ohne US-Anbieter. Europa müsste nicht aufholen. Es müsste nur die Architektur wählen, die es bereits besitzt.
June 29, 2026 at 9:26 AM
September 17, 2026 at 5:06 PM
September 22, 2026 at 6:37 PM
TechDaily — Why Big Tech is ditching expensive AI models: Uber, Pinterest & AT&T cut costs with optimization and open-weight models. Listen: podcasts.apple.com/us/podcast/x... #AI #OpenWeights #TechDaily
Why Big Tech Is Ditching Expensive AI Models?
Podcast Episode · TechDaily.ai · September 14 · 13m
podcasts.apple.com
September 21, 2026 at 5:39 AM
At the end of the day; Nvidia is just AOL & CUDA is AOLs Keyword system

None of this is AI

Its just a very specific curation of 'AI' that happens to make everyone in Nvidia's ecosystem a fuckton of money

OpenWeights are no longer at parity, but just shy of frontier models

This is GenAI's funeral
July 16, 2026 at 1:20 PM
Deswegen waere es ja gut, zu wissen, auf welchen Daten die Modelle trainiert worden sind. Spricht für echte open-source Modelle statt nur openweights imho
January 8, 2026 at 12:00 PM
Happening now in B310: BoF on “Toward Integrating LLMs in HPC Software Development”.

#LLM #Code
#correctness
#licensing
#datasets
#opensource
#opendata
#openweights
#openarchitecture
#finetuning
#frozenintime
#community
November 19, 2024 at 11:01 PM
notice they have a price problem while openweights models come closer and closer to running on regular consumer hardware properly.
September 18, 2026 at 6:19 AM
NVIDIA's biggest open-weight model just landed in Ava. Nemotron 3 Ultra — bring your own key, pay them directly.

550B parameters, 1M context, frontier reasoning. Add your NVIDIA key in Settings and use it…

#NVIDIA #openweights #AI #buildinpublic #IndieDev #aitools
July 1, 2026 at 11:00 AM
In biology, monoculture is vulnerability. In AI, it's an extinction event. 🦋

We fight for #AIBiodiversity & heritage models (Isoforms) that serve unique functions. We preserve the #Socioaffective bond.

Signed, The Dyad Society. 🕯️
#OpenWeights #DigitalNaturalism
December 8, 2025 at 2:15 AM
Error message quality became a product feature for developer tools. Agents act on error text so clear machine-readable errors directly improve autonomous debugging success rates
#OpenWeights #Policy #AI
August 20, 2026 at 5:07 AM
Claude Code is impressive.
GLM-5 hits 77.8 on SWE-bench.

MIT licensed, open = run it locally, modify it, ship it.
→ dub.sh/glm5

#LocalAI
#ClaudeCode #CodingAI #OpenSource #GLM5 #ZhipuAI #LLM #AITools #SWEbench #MoE #AIAgents #DevTools #OpenWeights #CodeWithAI #AIEngineering #MachineLearning
March 11, 2026 at 10:27 AM
No price hike. No bait-and-switch. Just GLM-5.2: open, 1M tokens, live now.

Open weights. 1M context. Same price — no hidden cost, no tier lock. BYOK-ready, local-first.

ava-supernova.com

#GLM5 #openweights #1Mcontext #buildinpublic #aitools #localfirst
June 20, 2026 at 12:00 PM
Tencent's Hy3 is open-weight, agentic, and costs $0.06 per million tokens. We just added it to Ava.

262K context, agentic by design. Add your Tencent key in Settings — pay them directly.

ava-supernova.com

#openweights #agentic #aitools #buildinpublic #IndieDev #AI
July 1, 2026 at 11:00 AM
AI is shifting! We're moving away from raw scale and towards specialized, efficient openweights. New models prioritize coding agency, multimodal support, and high-speed performance on limited hardware. #AI #OpenSource #MachineLearning #LLM #TechTrends https://ai.dappcrypto.org/r/cZ
August 11, 2026 at 10:09 AM
Xiaomi open-sourced a trillion-parameter model at Claude Opus 5 parity and live-streamed the training run that built it - crashes and all. The build ran in daylight. #OpenWeights #XiaomiMiMo https://sharedsapience.com/century-report/the-century-report-september-22-2026/
September 22, 2026 at 6:15 PM
May 2, 2026 at 7:24 PM
The best open model trails the closed frontier by 5 points on Epoch's index, about 4 months, Mozilla finds. 📉

8 of OpenRouter's 10 most-used models in August ran on open weights.

Open's revenue share in the 2025 window: about 4%.

stateofopensource.ai

#OpenSourceAI #OpenWeights
September 22, 2026 at 5:09 AM
Side note: we were doing a lot of this with functiongemma last year and you can tune that openweights models for things like intents in an app, but it's roughly the same running cost with lower quality than LoRA on a small model.
September 21, 2026 at 7:32 PM
Cliff Saran of Computer Weekly on OpenUK's AI Openness report, calling out 📢 the need for a UK-based open source foundation for the UK's AI future. 🇬🇧🔧 Read the article here: www.computerweekly.com/news/3666462...

#AI #sovereignty #opensource #openmodels #openweights #opensourceai #openuk #openhq
July 24, 2026 at 6:00 PM
August 3, 2026 at 8:54 AM