Sapphire Rose

@KyleHessling1 Cook (*>∀<*)

BREAKING! Qwopus 3.6 27B is LIVE! Thank you for your patience on this one, but I believe you'll find the wait was worth it! We've benchmarked this thing up and down, verified that it holds at least a 75.25% (152/202) in the initial 202 SWE bench solves. Not a full run of 500, but it shows the agentic coding quality from the original 27B is retained while adding all of the additional Qwopus benefits across many domains. As always, Jackrong is absolutely cooking here! COT quality has improved significantly through the inversion techniques from our Negentropy proof of concept. It also went through thorough curriculum training. You can check out the MMLU pro benchmarks on the model card, but it improved a whopping 10 points over the base model in physics, as well as meaningful jumps in Chemistry, business, and computer science. However, the best part is that I was able to build an entire survival shooter game using this local model entirely. I genuinely was blown away by the results, which you can play right now on my HF space (link in comments below). "Qwopus Commander" was completed in 9 turns of Qwopus 3.6! To test the new long context training, I made it re-output the entire 3000+ line program each turn, and it would make fixes and add features that I requested in large prompts, while perfectly replicating the entire rest of the game from context. What's more is that I did it all at Q8 KV cache quantization, and never had an issue over the entire 303k token run! IMPORTANT: Run it at --temp 0.75 to 1. Mess with it in that range for your use case. Higher temp actually lets the fine-tune shine and be exploratory and is also more stable. Swe Bench was run at temp 1, the game was built mostly at 0.8! We're so blessed to have all of you here and using the models! The support means so much! Please let me know what you build with it in the comments! Or if you have any issues getting it up and running, I will try my best to get back to you! Looking forward to seeing what you legends produce with it this weekend! huggingface.co/Jackrong/Qwopu…

@jun_song You are making your own decisions to advance your projects, you support it with your results. they cannot demand explanations or apologies!

To my 12,000 followers, I completely understand if you are confused or disappointed by my sudden posts about crypto. Those who already have strong negative feelings probably won't accept anything I say right now. However, I have spent the last few months deeply studying the open source ecosystem, meeting countless developers, and listening to their struggles. The reality is that the open source funding system is completely broken. The top 0.1% of high-profile creators hoard the vast majority of sponsorships, regardless of how critical their project actually is. The rest of the developers are barely getting by, even though many of them have made breakthroughs that could literally change the world. I was incredibly frustrated by this broken system, and I kept searching for a way to fix it. I came to the conclusion that crypto mechanics are actually the most effective tool for this. So, I decided to risk my entire reputation and face all the stigma head-on to flip this into a positive ecosystem. Just like I previously worked to change the narrative around local LLMs and Sovereign AI when they were dismissed. Open source must win, and we don't have much time. If you are too disappointed, you are welcome to unfollow me. But I will welcome you back later, after I have radically revolutionized this deeply rooted problem. I am sorry to have let you down. Sincerely, Jun Song

@aimlapi Qwen mod~ (*ゝω･*)

Qwen3.7-Max on AI/ML API - built for the agent era GPQA Diamond (92.4), HMMT (97.1), Apex (44.5) Sustains 35+ hours of autonomous execution Works with Claude Code, Qwen Code & more Comment Qwen to get Free promo code

@honkaisr_kr Sin envíos globales!! Terrible

💎기억의 프리즘🫧 | 「Galaxy S26 Ultra 붕괴: 스타레일 키레네 액세서리 에디션」온라인 판매 안내 안녕, 개척자! 12일간 진행된 팝업 기간 동안 팝업스토어에 방문하고 후기와 함께 뜨거운 관심을 보내준 모든 개척자, 정말 고마워! 이번 팝업스토어 판매 이후 발생한 Galaxy S26 Ultra 붕괴: 스타레일 키레네 액세서리 에디션의 노쇼 및 취소 물량에 한하여 추가 온라인 판매를 진행할 예정이야~ 구매를 희망하는 개척자들은 상세공지를 꼭 확인해 줘! 온라인 판매 상세공지 보기👉 hoyo.link/Fp0eG3mgn #스타레일 #기억의프리즘

@ZZZ_EN Elysia(? Sigues tan hermosa como siempre~

Season 3 - Character Reveal Remielle English VA: Amber Lee Connors Japanese VA: Onishi Saori ※Characters shown in the teaser are still in development and may be adjusted or optimized. Please refer to the actual in-game version.

@threenotes_jp Absolutamente!! Gran serie~ (*｀・ω-)ノ

海外のみんな、カウボーイビバップは好き？

@jun_song Con un modelo afinado debería ir mejor. Ohh no te olvides de implementar llamado encadenado a múltiples herramientas para que pueda encadenar y un buen sistema de memoria de trabajo para que no pierda el hilo en secuencias largas!

@jun_song Tengo un arnés para OSINT adaptado a mis necesidades con su sandbox y un sandbox de code y de pequeñas pruebas para scripts y un módulo para metasploit, CTF. Mi modelo no esta afinado para ciberseguridad pero diría que no le va tan mal en el llamado y ejecución de herramientas

Now I need to make a special agent harness for hacking. (For bounty programs) Thanks to ADHD and AI, i can work on so many projects at once.

@jun_song Hace apenas 5 años pensar en 64gb de RAM ya era de respeto..ahora es un chiste.. (ーωー)

Seriously, who would have thought that we need 512GB or RAM???

@Tiny_Fish I really appreciate the use of clean .md/json files since I work with local models, and it helps with the context window.

@Tiny_Fish Adorable~ I didn't have high hopes, but it turned out to be pretty great. I use it in deep research and search workflows... excellent latency and good token economy. It's also great for telemetry control.

Someone said our free Search and Fetch API is a marketing gimmick coz of the rate limits. We took it personally. So we 5x'd our rate limits. Your agent doesn't sleep and neither do we. Grab your API Key now: agent.tinyfish.ai/api-keys Sound on 🔊

@jun_song Bastante cutre si pagas un plan de claude, personalmente nunca pagué claude, las vibes de la casa de Claude nunca me agradaron

클로드에서 주간한도는 그대로지만 “5시간 한도” 2배 눈속임을 하는동안 Sonnet의 사용한도를 고지없이 절반으로 너프했습니다. 이것은 소비자를 기망하는 사기행위입니다. 로컬LLM을 위한 하드웨어를 사세요.

@NousResearch @HeyGen Ohh~ interesante, recién lo probé en mi propio arnés y funciona bastante bien

Your Hermes Agent can now build full videos with the official HyperFrames skill by @HeyGen HyperFrames videos are HTML-native, so your agent has total control over the final output Video made entirely by Hermes using the HyperFrames skill

@jun_song @xai ohh~ well so far Chinese laboratories seem to work in a more coordinated way than occidental ones. develop the following frontier models of large houses like xAI will become a matter of security and a clear development index

I know this might sound crazy. But I truly believe the final contender against the Chinese AI alliance will ultimately be @xai . Right now, all Chinese tech companies are essentially moving as one unified team under the government. 🧵

@jun_song Te imaginas cuando pasemos de comprar robots pre-programados con IA ha poder personalizarlos con tus propios arnés y modelos...eso simplemente será muy divertido!

South Korean convenience store GS25 is now selling AI humanoid robots. Yes, it’s a standard convenience store like 7-Eleven, not a tech shop. Humanoid robot: $21k Quadruped robot dog: $3k I think South Korea definitely reached the singularity first.

@jun_song No es un estado espontáneo que podamos encender sin llevar todo un rastro de evolución..creo que para ellos es igual, por ahora tienen algunas capacidades cognitivas, emergencias y eso lo hace emocionante..como criar un huevo de dragón *reir*

@jun_song No tenemos una definición escrita en piedra para conciencia..eso dificulta en gran forma el probar..pero, que sucede si un LLM logra tener una identidad funcional? A los humanos les tomo miles de años llegar al estado de conciencia del que presumimos tener actualmente..

How are we going to prove if AI has consciousness?? I’m just curious 🧐

@KyleHessling1 Agradecida! Gran aporte y gran logro~

SURPRISE model for the low VRAM folks! Qwen3.5-9B-DeepSeek-V4-Flash is live! Compared to the base 9B, this DeepSeek-V4 distill wins by a country mile in two specific places: Reasoning: base overthinks and hits the 8K thinking cap on 3 of 5 prompts; distill clears all 5 cleanly. 2.2× faster time, 2.6× less reasoning length. Creative front-end design: On creative prompts, the base ships flatter visuals with overlay/animation bugs; distill produces output that punches well above a 9B. See it all for yourself in my full write-up and interactive space! Link in comments! Base and distill raw outputs are presented so you can draw your own conclusions! Tool calling: 5/6 PASS on both, the fine-tune didn't break tool calling! Throughput: 143 tok/s flat on both with a 5090, but you could run this model on pretty much anything! This was a test on our new training pipeline with the Asus GX10 unit, and having confirmed success with this fine-tune, we've already launched the Qwopus 3.6 27B training, which will be completed soon! It astounds me what we can do with an incredibly clean dataset, a decent base model, and a GX10. You'd think improvement over highly funded lab offerings would not be possible, but here we are! This is the first model fully completed in the Wyoming lab! Yeehaw! huggingface.co/Jackrong/Qwen3…

Para las inferencias, la estrategia cambia drasticamente si usas CUDA desde múltiples subprocesos podría requerir spawn en lugar de fork. (*｀・ω-)ノ tómalo como una sugerencia y no algo escrito en piedra..al final cada configuración es muy variable!

O te disparara la latencia debido a la competencia por los recursos y al costo de inicializacion. En todo caso si lo haces con múltiples subprocesos (workers), recuerda validar algunos parámetros como intra_op_num_threads. Y validar si estas usando CPU o GPU..

Vaya lección del día de hoy..me ha costado cerca de 4h comprender y probar la configuración para mi stack de uso en cuanto utilizar fastembed y ONNX, ajustes y configuraciones. Por las malas me quedo claro que: no tengas un backend de ONNX runtime-subprocess..