twoimo @twoimo

Qwen 3.8 27b를 쓰시는 분들은 이 이미지 저장해 두시면 좋겠습니다. - 다들 Q4_K_S 많이 쓰실 거 같은데, 여유가 있어 램 용량 4G 정도 더 쓰면 BF 16기준 정확도를 약 3% 더 끌어올릴 수 있습니다. - Kv 캐쉬는 Q4가 아니라 Q8을 쓰면 속도와 정확도의 균형을 잡을 수 있습니다.

WebGPUで作ってる海 波打ち際が綺麗

Qwen3.8-27B holds 100% coherence to 256K on Nativ. 🚀 → Up to 933 tok/s prefill → 33 tok/s decode → Still 171 tok/s prefill and 11.5 tok/s decode at 256K → Peak memory 19.8GB at 512, 68.2GB at 256K No cloud. No API bills. Nothing leaves your Mac. Available now in Nativ v0.3.1

Qwen 3.8 35B A3B almost confirmed! Spotted on github! while everyone's busy swapping qwen3.6 27b for the new qwen3.8 27b dense drop on their single rtx 4090s and 3090s, the real quiet update is buried in this github commit: 35B total, ~3B active. same MoE recipe that made qwen3.6 35b a3b a favorite for 8-16GB VRAM setups. not officially announced yet, just sitting in the supported models table. if the pattern holds, weights drop soon. gpu poor gang, this is your sign It has been spotted at github.com/modelscope/ms-… This commit (ab726e9) in the modelscope/ms-swift repository is associated with the addition of support for the Qwen3.8 model family

Ever wanted an AI that doesn't hold back? Meet Qwen3.8 27B Heretic Abliterated, a GGUF quantized model that's been 'abliterated' to remove safety filters. It's uncensored, raw, and perfect for those who want unfiltered text generation. 5k downloads and counting!

Introducing GLM-5.3: Built to Code. Ready for Cyber Defense. - Top-tier coding and agentic capabilities, achieved through post-training on the 743B base model - A major leap in cybersecurity, setting a new standard among open models Tech Blog: z.ai/blog/glm-5.3

Gemini 3.7 Flash is now live in Google Antigravity 🚀 Our most intelligent workhorse model yet for coding and agents. Download or upgrade Antigravity to try 3.7 Flash today.

We’re launching DeepSeek-V4-Pro today! 🚀 🔷 Major Agent upgrades with strong production gains! 🔷 Flexible reasoning effort for V4-Pro & V4-Flash: low for simple tasks, high for daily Agent workflows, max for complex tasks. 🔷 Native OpenAI Responses API support, optimized for Codex with one-click setup. V4 Pro is now available on app/web. Try it via “Expert Mode”. V4 Pro is also available via API. Model names remain unchanged—please refer to the API docs for setup details.

We promised open weights for Qwen3.8. Now, time to meet them! 🎉 ⚡ Qwen3.8-27B: - A native multimodal dense model. With just 27B parameters, it outperforms Qwen3.7-Plus overall and shines in real-world coding & office workflows. - 262K native context, easily extendable to 1M tokens via YaRN. - Built for builders. Highly efficient, high-quality, and licensed under Apache 2.0. 🚀 The open weights for Qwen3.8-2.4T-A95B (Max-level) have also been released recently. Whether you're shipping lightweight applications with Qwen3.8-27B locally or building agents with Qwen3.8-2.4T-A95B, they're yours now! Download, deploy, and build something we haven't imagined yet. 👀👇 - Hugging Face： huggingface.co/collections/Qw… - ModelScope： modelscope.cn/collections/Qw…

Releasing the model weights and technical report of Kimi K3. Kimi K3 is our most capable model: a 2.8T MoE model with native visual understanding and a 1M-token context window. New model architecture: 2.5x the intelligence per unit of compute, not just more params. Alongside Kimi K3, we're opening up more of the stack behind it — high-performance attention kernels, MoE communication library, and infrastructure for running agent environments at scale. Model weights: huggingface.co/moonshotai/Kim… Tech report: github.com/MoonshotAI/Kim… Tech blog: kimi.com/blog/kimi-k3

Introducing Claude Opus 5. It's a thoughtful and proactive model that comes close to the frontier intelligence of Fable 5 at half the price.

@thsottiaux Congrats on 10M!@, but I messed up and used my banked reset, I'm suffering and crying. T_T.

Qwen3.8 is launching and going open-weight soon!🌐 With a massive 2.4T parameters, this model is continuously evolving. We believe it’s one of the most powerful model available today, compatible to leading frontier AI models , second only to Fable 5. You don't have to wait to test it. Just now, the Qwen3.8-Max-Preview made its debut on Alibaba’s Token Plan, Qoder, and QoderWork. Be among the very first to try it out. Can't wait to hear what you build. Stay tuned! 🚀  Token Plan international：qwencloud.com/pricing/token-… China：platform.qianwenai.com/pricing/token-…

@thsottiaux This would also give OpenAI valuable data to analyze and make better decisions later. It’s definitely worth a shot.

@thsottiaux In my opinion, just let users set their own limits. It’s the best way to respect our time. Of course, a warning should pop up, like: "Warning: Unclear tasks or system errors/updates can drain your tokens quickly. Do you still want to disable the 5h limit?"

@thsottiaux My suggestion might not be the perfect answer, but the friction is real. Why restrict everyone to a strict 5h box? Let users configure their own limits and find their optimal workflow.

@thsottiaux Right now, some people are forced to manage multiple accounts or jump to Business/Enterprise APIs just because of the 5h limit—which is a huge waste of efficiency.

@thsottiaux banked reset. please..

GPT-5.6와 Codex의 좋은 점은 복잡한 코드베이스에서도 요구사항→구현→검증 흐름을 이어갈 수 있다는 것입니다. AI 제품과 에이전트 기반 개발 워크플로를 실제 프로젝트에 적용하며, 데모가 아닌 유지보수 가능한 제품으로 만드는 과정을 탐구하고 있습니다. #GPT56 #Codex

@thsottiaux I manually reset around 12:30 AM on July 13th, but around 3:00 AM, my 5-hour limit disappeared and usage auto-reset. Had I known a full reset was coming, I wouldn't have used mine. Am I eligible for the banked reset?