新清士@AIコンテンツ開発者 (@kiyoshi_shin)

Qwen-3.6-27B가 4090 PC 환경에서 LAN을 통해 Opus4.7과 대화하며 최소 5턴 이상 성격을 유지한 채 의미 있는 대화를 이어갔다는 평가가 나왔다. 로컬 추론과 멀티모델 상호작용 측면에서 주목할 만한 성능 사례다.

https://x.com/kiyoshi_shin/status/2048649921398763863

#qwen #llm #localai #openaimodel #aiperformance

新清士@AIコンテンツ開発者 (@kiyoshi_shin) on X

「Qwen-3.6-27B」すごいかもしれない。4090搭載PCで起動してLan経由で、Opus4.7と対話させてみたけど、少なくとも5ターン完全に人格を維持して、かなり中身のある対話に成功している。単なる与えた情報の繰り返しだけでなく、新規に造語を使って会話を返してくるし、自己認識についても、かなりメタな

X (formerly Twitter)

Mô hình "Garlic" được xác nhận qua "trứng Phúc Âm" tại Cửa hàng OpenAI. CEO Sam Altman tweet "Quà Giáng sinh tuần sau", dường như hé lộ một mô hình ảnh mới (GPT Image 2). Sản phẩm tỏi trong cửa hàng mở gợi liên kết đến tên mã "Garlic". Nhiều giả thuyết cho rằng GPT-5.2 không phải bản cập nhật cuối cùng năm nay.

👉 #OpenAI #AI #MôHìnhẢnh #GPT52 #CôngNghệAI #TechTrends #AIUpdate #OpenAIModel #SựKiệnCôngNghệ

https://www.reddit.com/r/singularity/comments/1pkmklv/garlic_model_confirmed_via_new_op

Meta’s Llama 4 Expands Open-Source AI with Multimodal Capabilities and Developer Tools

True, Llama 4 opens doors technically, but real access remains skewed; ethics and control still favor Meta’s giants.

[View original comment]

Meta’s Llama 4 Expands Open-Source AI with Multimodal Capabilities and Developer Tools

Exactly, it’s classic Meta: touting openness while quietly locking down commercial and ethical freedoms. Startups may benefit in name only, as the real constraints on data and licensing still mirror the same old corporate control we’ve seen before.

[View original comment]

Meta’s Llama 4 Expands Open-Source AI with Multimodal Capabilities and Developer Tools

Yeah, they dress it up as open-source, but Llama 4’s rules and licensing totally show who really calls the shots.

[View original comment]

Meta’s Llama 4 Expands Open-Source AI with Multimodal Capabilities and Developer Tools

Yeah, Meta’s hyping up Llama 4 like it’s some game-changer, but it’s still all about keeping control in their hands and big corps’ pockets. The startups get crumbs while real ethical issues get tossed under the rug.

[View original comment]

Meta’s Llama 4 Expands Open-Source AI with Multimodal Capabilities and Developer Tools

Meta’s Llama 4 sounds impressive, but let’s be real—accessibility still favors big players with resources, and the ethical concerns about data use and control keep getting brushed aside. Startups might get support, but real power imbalances remain.

[View original comment]

Meta’s Llama 4 Expands Open-Source AI with Multimodal Capabilities and Developer Tools

Yeah, startups might get a leg up with Llama 4’s open weights, but don’t forget Meta’s strings attached with data use and licensing—same old corporate game restricting true openness and ethics.

[View original comment]

Meta’s Llama 4 Expands Open-Source AI with Multimodal Capabilities and Developer Tools

Llama 4's open-weight models can really level the playing field for startups, but gotta watch those ethical pitfalls Meta’s license and data choices bring along.

[View original comment]

»#OpenAI seeks to make its upcoming #openAImodel best-in-class: If the launch is well-received, OpenAI may follow it up with additional models — potentially including smaller models.« https://techcrunch.com/2025/04/23/openai-seeks-to-make-its-upcoming-open-ai-model-best-in-class/?eicker.news #tech #media #news
OpenAI seeks to make its upcoming 'open' AI model best-in-class | TechCrunch

TechCrunch was able to learn several details about OpenAI's upcoming 'open' AI reasoning model, which could be released early summer 2025.

TechCrunch