Mia ✓
@MiaAI_lab · joined 21 Jul 2022
Building with AI & LLMs | Insights, recipes, tools & honest experiments
30 636Followers
406Following
10 451Posts total
97.4KViews on collected posts
Postingan terbaru
Mencari postingan baru…
@MiaAI_lab Averaging 252 t/s for the last four days. https://t.co/V9Qh4PpL6y
569 views · 9 likes
· 0 reposts · 2 replies
Open on X →
And yes, I still prefer running it locally.
2.5K views · 26 likes
· 1 reposts · 1 replies
Open on X →
DeepSeek v4.1 Flash through API costs less than $3 for over 500 Million tokens.
Even more crazy: I measured a peak of ~315 tok/s decode and ~52k prefill https://t.co/TuKnmlNtGr
24.5K views · 548 likes
· 25 reposts · 60 replies
Open on X →
@MiaAI_lab I get nothing like these speeds, 30-50 tok / sec on a pair of gx10s
175 views · 1 likes
· 0 reposts · 0 replies
Open on X →
@MiaAI_lab I completely agree with Mia. And a lot of the time, low effort on Qwen3.8-Flash-Next works much better than xhigh. That’s especially true for turning data into JSON, in many cases, low actually gives me a higher accuracy rate than xhigh.
241 views · 2 likes
· 0 reposts · 0 replies
Open on X →
@MiaAI_lab it actually doubles the numbers on single spark 🫡
475 views · 1 likes
· 0 reposts · 2 replies
Open on X →
That's true.
Qwen3.8 Flash Next is the fastest 'flash' model you can run on 2x DGX Sparks
Get it here:
https://t.co/8WZdo8pBKi
Also, try sparkDash if you haven't already
https://t.co/2wBc5cB1Ay
22.2K views · 166 likes
· 11 reposts · 12 replies
Open on X →
For speed, nothing beats Qwen 3.8 Flash Next on 2x Sparks. Benchmark for code pushing 110 tok/s c=1 and up to 688 tok/s on c=16 https://t.co/TRKSwn3L4O
19.9K views · 28 likes
· 1 reposts · 10 replies
Open on X →
Multi GPU bridge that allows AMD GPUs to work together on Windows with vLLM. Tested on AMD R9700 AI Pro cards.
Check it out if you have AMD GPUs 👇
496 views · 4 likes
· 0 reposts · 0 replies
Open on X →
I have been working on this multi-gpu windows bridge for my AMD windows port of vLLM and ROCmFPX. It brings Direct-RCCL on Windows with D3D12 and hybrid bridging. It's not Peer to Peer but it's the closest we can get to it. When I started this, there were none in existence like
554 views · 10 likes
· 1 reposts · 2 replies
Open on X →
We're making AMD great again https://t.co/6B3RjKq7eB
1.8K views · 16 likes
· 2 reposts · 1 replies
Open on X →
@MiaAI_lab @MiaAI_lab that's awesome! curious how it handled color palettes. found some AIs tend to pick flashy combos that aren't always user-friendly. did you notice any of that?
123 views · 2 likes
· 0 reposts · 0 replies
Open on X →
@MiaAI_lab What! 😅 better than Astra!? Im going to tell my agent you said that 🙈😂
326 views · 2 likes
· 0 reposts · 1 replies
Open on X →
@MiaAI_lab I have been waiting for this one.
424 views · 3 likes
· 0 reposts · 1 replies
Open on X →
I asked DeepSeek v4.1 Flash to create 100 HTML files
Rules were simple:
- look stunning
- zero repetitive designs
- go full creative mode
The purpose of this test was to evaluate the model's creativity alongside its frontend design skills.
The model is clearly better at
11.3K views · 139 likes
· 8 reposts · 18 replies
Open on X →
@MiaAI_lab This!! Yes. GLM 5.5 could be our Fable..! Way to go.. @NVIDIAAI you watching.. need more sparks over here for the @MiaAI_lab consortium.. I am considering more sparks. I will match one for one each you give the lab..!
272 views · 4 likes
· 0 reposts · 0 replies
Open on X →
This is THE update everyone was waiting for ⚡️
MASSIVE speed gains for GLM 5.3 Flash EXL3 running on 2x DGX Sparks
- 29% faster on average for prose (!)
- 20% faster on average for code
Prose decode speed is now around 36 tok/s!
This is the FASTEST GLM 5.3 Flash setup yet ht
9.3K views · 255 likes
· 17 reposts · 36 replies
Open on X →
Dibandingkan akun berukuran sama
20 postingan dari 90 hari terakhir, dibandingkan dengan rentang 10K–100K pengikut. di bawah akun sekelasnya baik pada jangkauan maupun interaksi.
Median tayangan525akun ini924median untuk 10K–100K
Jangkauan, %1.71%akun ini3.62%median untuk 10K–100K
Interaksi, %0.97%akun ini1.52%median untuk 10K–100K
| Metrik | Akun ini | Median untuk 10K–100K | Rasio |
|---|---|---|---|
| Median tayangan per postingan | 525 | 924 | 0.57× |
| Jangkauan (tayangan ÷ pengikut) | 1.71% | 3.62% | 0.47× |
| Tingkat interaksi | 0.97% | 1.52% | 0.64× |
Akun lain pada rentang ini → Bandingkan dengan akun lain → Bagaimana tolok ukur ini disusun →
Growth & engagement
How the posts we collected actually performed: views and reaction rate post by post, what the audience did with them, and where the follower count goes.
Views per post
3218 Sep
123
1.8K
1.5K
554
496
19.9K14 Sep
22.2K
475
241
175
24.5K15 Sep
2.5K
569
Last 14 collected posts, oldest on the left. The scale is logarithmic: one post can outrun the rest a hundred times over.
Engagement rate per post
0.93%8 Sep
1.63%
1.10%
1.00%
2.53%
0.81%
0.20%14 Sep
0.86%
0.63%
0.83%
0.57%
2.58%15 Sep
1.14%
1.93%
Reactions — likes, reposts, replies and quotes — divided by views. Median for 10K–100K accounts is 1.52%.
What the audience does
Likes66.7%1 233 in total
Reposts3.6%66 in total
Replies8.0%148 in total
Quotes0.8%14 in total
Bookmarks20.9%387 in total
Share of every reaction we collected for this account. Replies mean argument, reposts mean endorsement, bookmarks mean the post was worth keeping.
The follower curve appears once this account has two daily snapshots — we take one a day, and this one is on its first.
Akun serupa
silvan ✓
@silvanrec
61.3K followers
· 1.9K posts
すでたき ✓
@sudetaki
61.3K followers
· 13.2K posts
The Press Democrat ✓
@NorthBayNews
61.3K followers
· 161.8K posts
Rangers Women ✓
@RangersWFC
61.3K followers
· 24.7K posts
Jamal Baloch
@Jam__Baloch
61.3K followers
· 50.8K posts
Picoty ✓
@PicotyChelangat
61.3K followers
· 55.2K posts
Men Life Legacy ♠️ ✓
@menlifelegacy
61.3K followers
· 3.2K posts
EtFelicitoFill ✓
@etfelicitofill
61.3K followers
· 99.3K posts