The discussion centers on whether Edge0-145B is a genuine architectural innovation or a derivative of Qwen Flash Next, with users debating the validity of its open-source status and hardware compatibility claims. While some participants express excitement about local deployment capabilities, others criticize the restrictive licensing and lack of clear instructions for non-NVIDIA platforms.
27 readable replies · Supporting reply 1 · Supporting reply 2 · Supporting reply 3 · Supporting reply 4Niko1221
Strata
C++2026–2026
Niko1221 / Strata
Strata
Qwen3.8-Flash-Next on any consumer hardware: one-click install for Windows / Linux. Strata inference engine, OpenAI/Anthropic API on localhost, optional image input.
17Distinct people
19Mentions
0Explicit recommenders
43.5KGitHub stars
What to know before exploring
What it is
Qwen3.8-Flash-Next on any consumer hardware: one-click install for Windows / Linux. Strata inference engine, OpenAI/Anthropic API on localhost, optional image input.
GitHub
Where it is used
- Self-host Qwen3.8-Flash-Next (SSD Nodes guide) — Self-host Qwen3.8-Flash-Next uses Strata as an inference engine to run Qwen3.8-Flash-Next on consumer PCs. Source
- Strata Manager — Strata Manager uses Strata to run local inference and manage runtime setup, configurations and service lifecycle. Its capabilities depend on the installed Strata runtime. Source
What people are saying
No recommendation phrase was found in the indexed comments.
No classified warning is attached to this page. Read the discussions before treating silence as approval.
Also in these discussions
No other repository is repeatedly linked in the same discussions.
This week on Hacker News
Nov 7–Nov 14 UTC · Direct repository links and separately labeled related project stories. Indexed through 11/14/2026, 3:03:16 PM UTC.
Read the discussion on HN. A full-thread summary is not available yet.
The discussion centers on the viability of running heavily quantized large language models locally, with users debating whether sub-4-bit quantizations offer usable quality or render outputs incoherent. Participants also argue over the economic and practical justification for self-hosting hardware versus paying for cloud subscriptions, citing factors like speed, intelligence gaps, and data sovereignty.
209 readable replies · Supporting reply 1 · Supporting reply 2 · Supporting reply 3 · Supporting reply 4Mentions over time
HN mentions by week19 dated mentions
Observed mentions: 2026-09-28 15:07:39 UTC to 2026-10-31 07:36:05 UTC. Weeks start Monday, UTC.
Counts come from the repository’s dated source evidence. Index cutoff: 2026-11-02 08:16:59 UTC.
View mention counts and date ranges
| Period | Start (UTC) | End (exclusive, UTC) | Mentions | Submissions | Comments |
|---|---|---|---|---|---|
| Sep 28 | 2026-09-28 00:00:00 UTC | 2026-10-05 00:00:00 UTC | 6 | 3 | 3 |
| Oct 5 | 2026-10-05 00:00:00 UTC | 2026-10-12 00:00:00 UTC | 2 | 0 | 2 |
| Oct 12 | 2026-10-12 00:00:00 UTC | 2026-10-19 00:00:00 UTC | 4 | 0 | 4 |
| Oct 19 | 2026-10-19 00:00:00 UTC | 2026-10-26 00:00:00 UTC | 3 | 0 | 3 |
| Oct 26 | 2026-10-26 00:00:00 UTC | 2026-11-02 00:00:00 UTC | 4 | 0 | 4 |
What Hacker News says
19 sourcesSKskrellm·Oct 31, 2026
Yep, I agree. You can already run them locally on a top-shelf gamer PC with decent token throughput: https://github.com/Niko1221/Strata That's today.
As for tomorrow, consider that literally every chip maker is working on better and better NPUs to boost local performance (Ryzen AI, Intel Ultra, Apple M7, Broadcom PON, Huawei Ascend, Snapdragon X Elite, etc.), and imagine optimized models running on these, yeah, insanely huge potential.
As of 2026Q4 I wouldn't bet on any cloud AI, that's for sure.
In A tiny world model running games on a Raspberry Pi Pico 2 WI've benchmarked quite a few models on my 9070XT, and I basically gave up hope of using models locally: https://claude.ai/artifact/77pHgucqQNsmic8ZEnqdBA
I asked Claude to come up with 7 problems to solve that were not part of any benchmarks. Besides speed, I asked it to put together a scoring that would reflect models that perform well on the 7 problems and in a reasonable time-frame (RWS, the definition is in the artifact). I valued models that produced good answers in a reasonable time better than models that had very high throughput but kept spitting useless stuff. Overall, IIRC most of them still had a >30% chance of producing wrong/non-working replies.
I then tested Strata ( https://github.com/niko1221/strata ) and the results are not even close. You can basically ignore everything in that table and use Strata instead, you can use b
Read the rest
igger/less-quantized models and have way better results. Basically any otherwise-decent model degrades noticeably at Q < 5, and KV cache at Q4 also contributes a lot.
Still, a 20$/mo Claude subscription is way cheaper/more efficient if you code 2-3h in the evening than running stuff locally, unless you really have hardware than can hold models that get most things right at the first try. You can iterate way faster and cover a lot more ground in 2-3h using a Claude subscription, or OpenRouter if you prefer to change models. You get faster replies that are 99% correct at the first try, whereas using the local model you'd be spending 30% of your time fixing things that don't work and prompting it to do better.
Nowadays I use the local model to delegate tasks that would burn a lot of tokens, like sifting through big data archives for indexing and such.
If you have enough system RAM (64GB), you can run Qwen 3.8 Flash Next with a GPU as small as 12GB, this repo is pretty amazing:
https://github.com/Niko1221/Strata/
My setup is a Ryzen 3600, 90GB DDR4 allocated to a VM (don’t even really need that much, just had it available, special thanks to the year 2024), and an RX 9060XT 16GB ($500 GPU).
I get around 40-60 tok/s depending on the quantization level I choose.
As you can see, the CPU is nothing special, don’t even need DDR5.
In New iMac with M6Take a look at the docs for https://github.com/niko1221/strata for some ideas, particularly https://github.com/Niko1221/Strata/blob/main/docs/MODELS.md#... .
In OpenAI and Anthropic cut their LLM prices by up to 80% within a few months. Why?Any chance you could try qwen3.8 flash next? That is currently the best local model for actual coding tasks in my + many coworkers experience. It will run reasonably fast if you have 64+ gb ram and use strata or a strata fork (vram not super important with this method).
https://github.com/Niko1221/Strata
In MiniMax M3.1 FlashI ran the Strata project with this on 32gb ram 16gb AMD 7900XTX. It goes about 20tok/s off the ssd. I'm guessing 64gb would go to 40tok/s. https://github.com/Niko1221/Strata Quality wise its not fast enough for me to eval. I'll probably spend some tokens on it through openrouter. DeepSeek 4 flash though is a mess compared to GPT Sol 6.1.
In DeepSeek V4.1 Flash 1017 (Max)You can get something relatively capable and usable with your configuration and Strata: https://github.com/Niko1221/Strata
In DeepSeek V4.1 Flash 1017 (Max)A Strata (https://github.com/Niko1221/Strata) challenger!
The first one I've seen (although I'm sure there are and will be more as more time progresses...)
It seems to require a bit more system memory (70GiB free RAM as opposed to 32GB) and two newer nVidia RTX 50 series graphics cards with 16GB (as opposed to a larger choice of multi-vendor older single graphics cards with a minimum of 12GB), but on the flip side, it does seem to make up for that with faster token generation speed...
Anyway, it seems to be worth watching...
Strata may have a worthy competitor here!
In Show HN: Capy-flash – a 125B MoE model on two 16 GB consumer GPUsOh you should be running flash next with strata. Just go to stratas github page [0] and it will explain everything. It runs quite fast on 16+64, you should be able to get 50 tok/s or so. It's a 180b model and it feels a lot smarter than most of the other open weight options. As far as knowing "what will fit" in general that's hard to know these days. The moe models are hard to predict and run wildly different in different engines (llama.cpp, vllm, strata likes, etc) but they tend to be the only good options around 16gb vram. Dense models like qwen3.8 27b are only good if they're entirely in vram. Moe models you need the active count + a few experts (although it will run faster the more experts you can fit).
[0] https://github.com/Niko1221/Strata
In LFM2.5-24B-A2B: Built for fast on-device agentsThen check out those:
https://github.com/prabanta-dev/janas <- this, focuses on Qwen3-Next-80B-A3B.
https://github.com/Niko1221/Strata <- Qwen3.8-Flash-Next. I use this (tuned) and I cannot understate how satisfied I am.
https://github.com/antirez/ds4 <- DeepSeek 4 family + more.
In Janas: Run LLMs larger than RAM by streaming experts from SSDQwen 3? (Jumanji) What year is it?
What a waste of time.
Strata already does this with Qwen3.8-Flash-Next: https://github.com/Niko1221/Strata
In Janas: Run LLMs larger than RAM by streaming experts from SSDNot sure why this is downvoted.
It is fully possible to run the 180B Owen model on strix halo using strata and even get reasonable speed out of it.
https://github.com/Niko1221/Strata
In Step 5 Preview, a 1M-context MoE from StepFun, shows up on OpenRouterwhat they should freak out about is qwen 3.8 github.com/Niko1221/Strata
it rips with just 64 ram and a 9070xt
In Why isn't the industry freaking out about DeepSeek 4.1 Flash?> Set up Strata on this PC for me: https://github.com/Niko1221/Strata - follow docs/AI_SETUP.md in that repository.
And I thought piping to bash was bad
In Run Qwen 3.8 Flash Next (125B) on consumer hardware (RTX 4090) at 100T/sThey have some community benchmarks published https://github.com/Niko1221/Strata/tree/main/bench/results
In Run Qwen 3.8 Flash Next (125B) on consumer hardware (RTX 4090) at 100T/sShowing the 15 most recent of 19 sources. Enable JavaScript to search and browse every indexed mention.