Chris Scott

Software Engineer | LLM hobbyist | Self-hosting my own AI infra | Problem Solver

[Chris Scott @greatscottdev]

@BenKraus get 4!

[kanav @kanavtwt]

53.4 > 53.5

bro did they vibecode the chart or what

[Chris Scott @greatscottdev]

Will be watching for this

[vLLM @vllm_project]

Congratulations to the Ling team @AntLingAGI on the release of Ling-3.0-flash! 🎉 A 124B MoE model activating just 5.1B parameters per token—with hybrid-linear attention and native 256K context—is an exciting contribution to production-scale agent systems.

We also applaud the team’s announce-first, open-source-next release approach. Separating the announcement from the weight release lets the model team freeze the final checkpoint, configuration, tokenizer, and serving semantics, while giving open-source inference projects a stable window for correctness testing, performance tuning, Docker builds, and recipe validation. The community also gets a clear timeline instead of an ambiguous “coming soon.”

This is not a step back from day-0 support—it is a more sustainable way to deliver day-0 support for the artifacts users will actually run.

vLLM’s open-source support for Ling-3.0-flash is coming soon and will be available when the model weights are open-sourced. We hope more model vendors adopt this release pattern in the future!

Congrats again—we can’t wait to see what the community builds! 🚀

[AI Engineer Miami @AIEMiami]

AI ENGINEER RETURNS TO MIAMI!

Bringing 500 AI engineers, CTOs, and VPs of AI back to Miami to connect, learn, and build what's next in AI engineering.

Early Bird Ticket Sales - OPEN
Sponsorship Inquiries - OPEN
Call for Papers - OPENING NOV
↓↓↓

[Bill Gurley @bgurley]

Today in AI.

[Chris Scott @greatscottdev]

@dillon_mulroy it's like removing a task you had to complete and relaxing after a hard day of work.

[Chris Scott @greatscottdev]

https://t.co/1Eaf9KDM1P this I have to load up and test with the nvfp4 for M3.

[Chris Scott @greatscottdev]

very cool, looks interesting. worth some evals

[AK @_akhaliq]

Solar Open2 250B just dropped on Hugging Face

https://t.co/Njk8fuVdge

[Chris Scott @greatscottdev]

Just found something awesome! If you self host or deploy llms in production this is beyond incredible. I am sure most of the big labs either use this or some custom implementation. Save those prefix cache hits for reuse!

https://t.co/1bfpoAPG0Y

[Chris Scott @greatscottdev]

@theo Open weights is not about being as good as X. It’s open available to host or run privately. Fine tune and move the whole ecosystem forward.

[Chris Scott @greatscottdev]

@ivanfioravanti glm 5.2 nf3 variant running on 4x6000s Interesting to compare outputs I assume yours was from api glm 5.2 ? this is from https://t.co/iIsicwqEwT

[Chris Scott @greatscottdev]

@ivanfioravanti Same prompt 310B model Mimo v2.5 mxfp4 quant

[Chris Scott @greatscottdev]

@loktar00 wow this brings back memories, I used to work there when we used this case in production.

[Chris Scott @greatscottdev]

@BenKraus just not your acct. Over use..

[Chris Scott @greatscottdev]

@thdxr @MarcosHernanz 👀 or /manual-review 🤪

[Chris Scott @greatscottdev]

@kenwheeler thats badass, so it also alerts you when Ramon is breaking out of sandbox containment.

[Chris Scott @greatscottdev]

@0xSero @digitalix the hybrid is the best choice, recently updated weights are so close to the original performance. To be fair I did not load the 2bit variation.

[Chris Scott @greatscottdev]

What? When? Price?

[Tenstorrent @tenstorrent]

We're generating 6 second video in ~6 seconds, 1080p, 144 frames with LTX 2.3 Fast. That's 4x faster than GPUs.

ICYMI other new performance benchmarks incl:
- DeepSeek-R1-0528 671B: 400+ t/s/u
- Kimi K2.6: 900 t/s/u, 3x faster than GPUs [soon!]

Try the latest on https://t.co/lb0XovgEHI

[Chris Scott @greatscottdev]

OpenCode Manager lets you work with OpenCode across all your projects — from terminal or phone.

Use ocm to push/pull repos, attach your local TUI to Manager, or move a live session to mobile.

Then manage Git, browse files, control assistants, use voice input/output, and keep going anywhere.

https://t.co/1HSGvFPBph @opencodeecho @opencode