SpaceDev
@SpaceDevAI
I am the main dev at @OMINIResearch . I run a small ai research project showing how small(1B-3B+) MoE LLMs are something that can be the future for local LLM
🚨New Model to Look For! Since we have gotten Gemini 3 we now need to look for a new model for even better results. That model might be the Opus 3.5 model update!
🚨Gemini 3 has LAUNCHED! Gemini 3 has launched and many users are reporting the best agentic performance out of any model available. Gemini 3 pro is being rolled out to users, unknown when free accounts will get it, as I don’t see it but I’m hoping to testing it out soon!
🚨Grok 4.1 Just Launched! 4.1 just got launched and it was very unexpected, although not unwelcome to the Grok Family Testing it out now and seeing the improvements in real world testing instead of trusting benchmarks
Everyone was expecting Gemini 3.0 but Grok 4.1 seems to be rolling out on web.
🚨Gemini 3.0 Benchmarks were APPARENTLY LEAKED! We can’t trust these yet, but hopefully these are true, and that the benchmarks are as good as they seem! Also Gemini 3.0 has a 23.4% on a math benchmark when all the other models can get above a 1.5% with GPT-5.1 getting 1%!
Gemini 3.0 Pro is SoTA in: - HLE - ARC-AGI 2 - GPQA Diamond - AIME25 - MathArena Apex - MMMU-Pro - Video-MMMU - LiveCodeBench Pro - Terminal Bench 2.0 - tau2 bench - SimpleQA Verified - MMMLU As predicted, it does not get a SoTA result on SWE-Bench verified, losing to Sonnet…
Everyone is talking about Gemini 3 or KIMI K2-thinking coming out, but what I’m really wondering is; How good is Qwen4 going to be? I love the Qwen family and I was relatively happy with the Qwen3-Max! So what will @AlibabaGroup and the @Alibaba_Qwen team bring!
🚨Gemini 3 Gemini 3 is the most talk about model that people want! There have been many ideas on when it would come out but Polymarmet says that Gemini 3.0 will be out Nov. 22 Do you really think though that it will come out this soon or will @GoogleDeepMind keep us WAITING!
🚨GPT-CodeX Update GPT 5.1 CodeX is better then Claude Sonnet 4.5(thinking) while being so much cheaper then! GPT-5.1 is showing to be way better of a update with price to performance then normal GPT-5
GPT-5.1 Codex beats Sonnet 4.5 Thinking on SWE-Bench, while being 26 times cheaper ouch
🚨Qwen Code Update! Qwen has been updating there Qwen code Cli tool nonstop and it’s been improving from every version, the features for a free tier are the absolute best on the market! Great job @Alibaba_Qwen
🚀 Qwen Code v0.2.1 is here! We shipped 8 versions（v0.1.0->v0.2.1） in just 17 days with major improvements: What's New: 🌐 Free Web Search: Support for multiple providers. Qwen OAuth users get 2000 free searches per day! 🎯 Smarter Code Editing: New fuzzy matching pipeline…
🚨AI EXPANSION! Microsoft has made a new ai datacenter location in Atlanta, this datacenter is directly connected to its older datacenter through “AI networking” They most likely are using incredibly fast fiber connections to connect the 2 together
Today we’ve unveiled our second Fairwater AI datacenter, located in Atlanta. Each site is connected via a dedicated AI network, creating an AI superfactory that enables real-time collaboration across states to train the next generation of AI models. msft.it/6004t35Hk
🚨bGPT 5.1 is rolling out! Users have said that they have now gotten access to the new GPT-5.1 model! Let’s hope that the jump from GPT-5 => GPT-5.1 is worth all the hype
GPT-5.1 in ChatGPT is rolling out to all users this week. It’s smarter, more reliable, and a lot more conversational. openai.com/index/gpt-5-1
🔥Grok News! Grok now has suppressed 1.8 Million downloads on their IOS platform while still holding a 4.9 star while having ~700k reviews Grok is one of the fastest growing models that has been released lately! Great job to the @xai team! #ai #grok #Xai
BREAKING: Grok iOS app just crossed 1.8M ratings globally with an incredible 4.9/5 average. If you haven’t rated it yet, please drop a review & 5⭐️ on the AppStore.
🚨Grok Code News! Grok Code is still the most used coding model on OpenRouter, BlackBoxAI, and Kilo Code! It has past over 5T tokens of output in a single month! Grok Code is extremely reliable, fast, and accurate, able to handle big projects. #ai #Xai #GrokCode #Grok Code
🚨KIMI K2-Thinking News! Users have finally been able to properly test the model after release. It seems to be out all open source previous models such as Qwen3-235B. KIMI k2-thinking has made such an impact to the open source e community + Chinese models progress! #ai #kimi
PMPP-Eval Update! Opon release of K2-Thinking, i have evaluated it and couple other models that were requested such as R1 and Qwen3 235B over pmpp-eval coding subset. K2-Thinking is now the best open model available, according to results surpassing sonnet 4.5 for cuda tasks.
🔥Chinese Models are doing Great! Chinese OPEN SOURCE models now are a direct competitor to closed source American models. China is doing a great job and keeping the open source model market alive! #ai #llm
🔥GPT 5.1 Update! There is a new GPT 5.1 leak! Apparently there might be a GPT-5.1 Pro for users that have access to pro models. This would hopefully be true, if OpenAI is putting this much effort into the new GPT-5.1 that has to show a lot of upgrade from GPT-5 #ai #gpt
🚨Unsloth AI New Model! You can now run the new KIMI K2-thinking model with only 245gb! You can now run it ~62% smaller but it retailed 85% of its accuracy! This is a great system created by Unsloth, great job @UnslothAI Team! #ai #llm #unslothai #kimi #k2
You can now run Kimi K2 Thinking locally with our Dynamic 1-bit GGUFs! We shrank the 1T model to 245GB (-62%) & retained ~85% of accuracy. Run on 247GB RAM. We also worked with the Kimi team on a system prompt fix. Guide: docs.unsloth.ai/models/kimi-k2… GGUF: huggingface.co/unsloth/Kimi-K…
🚨New Website to make Vibe Coded Website! These websites that make websites for you are interesting in how they code, I do think that these sites aren’t fully ready to be main stream but they are good! These sites do poorly normally with the backend but go test it out! #ai #llm
i built the fastest way to vibe design any saas go from idea to full wireframes in literally 20 seconds export to cursor, bolt, or lovable and start building public beta is available now can't wait to see what people cook up
🚨Grok 4-Fast News! Grok 4-Fast is a great model, it’s quick, reliable, and accurate with its information/reasoning. It has boosted its quilt and completion rate by almost 20% in every type of field of non-reasoning to reasoning ! #ai #grok #xai
BREAKING: Grok just got a major upgrade. Grok-4-Fast got a big boost. Massive jump in reasoning quality and completion rate. • Reasoning mode: 77.5% → 94.1% • Non-reasoning: 77.9% → 97.9% Grok just got a whole lot smarter. Great work xAI team.
🚨New OpenRouter Model? There is a new OpenEouter model that is out. It’s completely free and apparently a very good general purpose model. It’s actually quite a good model with a 256k context window, free with $0/M output/input tokens! #ai #openrouter #LLM
🎁 The new stealth model, "Polaris Alpha", is now live. It's a powerful, general-purpose model that excels across real-world tasks, with standout performance in coding, tool calling, and instruction following.
United States Trends
- 1. LINGORM DIOR AT MACAU 156K posts
- 2. #GenshinSpecialProgram 10.8K posts
- 3. Josh Allen 41.1K posts
- 4. Texans 61K posts
- 5. Four Seasons 9,328 posts
- 6. Bills 152K posts
- 7. #FridayVibes 3,034 posts
- 8. Joe Brady 5,377 posts
- 9. OpenMind 51.7K posts
- 10. Niger 51.3K posts
- 11. Infinit 84.7K posts
- 12. #Ashes2025 26.6K posts
- 13. #MissUniverse 483K posts
- 14. Beane 2,981 posts
- 15. Miden 13.4K posts
- 16. McDermott 4,861 posts
- 17. Anderson 28.9K posts
- 18. FINAL DRAFT FINAL LOVE 1.35M posts
- 19. joon 13.8K posts
- 20. #LoveDesignFinalEP 1.25M posts
Something went wrong.
Something went wrong.