@Alibaba_Qwen @arena Yes if you actually going to release weights for 397b variant! This is what I am still running at home but stuck on 3.5
English
Chris Scott
292 posts
Chris Scott
@greatscottdev
Software Engineer | LLM hobbyist | Problem Solver
Florida Katılım Ekim 2023
153 Takip Edilen82 Takipçiler
🚀🚀Qwen3.7 Preview lands on Arena ！
Here come Qwen3.7-Max-Preview & Qwen3.7-Plus-Preview. Alibaba now #6 lab in Text, #5 in Vision.⚡️⚡️
Can't wait to release Qwen3.7 series models！Stay tuned! @arena
Arena.ai@arena
Qwen3.7 Preview By @Alibaba_Qwen lands on Arena for Text and Vision. In Text Arena, Qwen3.7 Max Preview ranks #13 overall. Alibaba is now the #6 lab in this arena. - #7 Math - #9 Expert - #9 Software & IT - #10 Coding In Vision Arena: Qwen3.7 Plus Preview ranks #16 overall, making Alibaba the #5 lab. Congrats to the @Alibaba_Qwen team on the latest progress!
English
@jahooma Missing @XiaomiMiMo is that to keep it in the lead?
English
DeepSeek v4 **Flash** is absolutely insane.
It costs almost nothing (~1/300th Opus), and yet performs among the best open source models.
On our coding benchmark Flash does better(!) than Pro
English
@code_barbarian @joinhello_miami @remix_run @mongoosejs Enjoyed the talk, learned some things I didn’t know about both @mongoosejs and @remix_run
English
Check out my talk on @remix_run + @mongoosejs in wynwood tonight 🌴
Remix Miami@remix_miami
Join us next week as @code_barbarian teaches us how to work with data in @remix_run 3 and what really matters! Join me at Working with Data in Remix 3: What Actually Matters | Remix Miami meetu.ps/e/Q1D3G/LpYmM/i luma.com/u6a3zdv3
English
ERNIE 5.1 just dropped.
Built on ERNIE 5.0's pre-training foundation, our latest foundation model upgrades search, reasoning, knowledge Q&A, creative writing, and agentic capabilities, while using only around 6% of the pre-training cost of comparable models.
More in the thread 🧵
English
Chris Scott retweetledi
shipped another worktree improvement: opencode will show all worktrees available from git
you don't have to create them through opencode anymore! use whatever tool you want to name and organize them
English
@mattpocockuk @techgirl1908 I think so for now. All the memory implementations I have used end up causing more problems than helping with the exception of a single loop to complete a single outcome.
English
@techgirl1908 I am currently intrigued but sceptical about memory, specifically about engineering.
Is it not better to optimise for the most common, predictable, cheapest state - no memory?
English
The more I work with agents, the more I'm convinced that "just give it more context" can't be the whole answer.
I'm not seeing enough discourse about memory. More specifically, memory design... like what gets stored, what gets retrieved, what gets summarized, what triggers the agent to look things up again.
I'll be spending time with @oracledevelopers soon, getting hands-on with agentic memory patterns. Very excited to get into the weeds!
English
@vllm_project @lightseekorg Sm120 support? Or only enterprise Blackwell*
English
🚀 Excited to be the exclusive day-0 launch partner for @lightseekorg's Tokenspeed project!
We've integrated Tokenspeed's MLA library, optimized specifically for agentic workloads with long context and multi-turn, purpose-built for Kimi 2.5/2.6 and DeepSeek R1 on NVIDIA Blackwell hardware!
Try it out today with our preview image - nightly support coming soon!
LightSeek Foundation@lightseekorg
Introducing TokenSpeed, a speed-of-light LLM inference engine. > TensorRT LLM level performance > vLLM level usability > Built by a lean and mission-driven team in two months > MIT license, open-source github.com/lightseekorg/t… lightseek.org/blog/lightseek…
English
Chris Scott retweetledi
@code_barbarian is going to give us insight to handling data in @remix_run 3!
Tuesday May 12th at @remix_miami located at the Dock Wynwood.
Come join us!! Sign ups below
English
people who don’t clean the community grill after use, are the lowest of the low
English
@zUnEm01 for me it come to Qwen 3.5 397b. It is what I currently run locally. I would love for the 3.6 to hit open weights. @Alibaba_Qwen I thought was all about OSS?? It offers the best in option for me in speed / Code / Instruction following out of other options I can self host.
English
GLM 5 could be better but it is very unreliable asf!
Kimi k2.6 could be better but it just has issues with understanding and following instructions, it over does stuffs and destroys my repo.
Deepseek is a winner here because it understands and follows instructions with 1m context it's a plus for me. The only problem with Deepseek is this: it doesn't have vision.
Kasif@md_kasif_uddin
Be honest, which is the best open source AI Model?
English
@0xSero I did give this a quick test. What did you run it on? Sliding window attention does not have the best support for sm120.
English
MiMo-V2.5
Supports: vision + video + audio
Speeds:
- 128 tok/s decode
- 367 tok/s prefill & 30K cache
- FP8 weights FP16 cache
- 65% on terminal-2
- 58% on swe-pro
- 62% on claw-eval
- 376gb VRAM
- 256k context
Tying or closely trailing Kimi-K2.6
Excellent work by Xiaomi
English
English
@rot13maxi @QuixiAI Not really, 397 is the most useful option.
English
What?? This is interesting. I do use warp, not for any of the agent features. But I was one of the ones pushing to allow for custom local / endpoints. (github issues)
Warp@warpdotdev
Warp is now open-source.
English
