Head Coach @ Precision Athletics Fencing 🤺 | Coaching national medalists - Foil 🏅 | Camp Director, Fencer's Paradise 2026 🌺 | AI Tinkerer 💻 | Climber 🧗
Los Angeles, CA
Joined October 2009
- Tweets2.4K
- Following260
- Followers251
- Likes5.2K
Wow, this will be my main model... At least until we see what Qwen releases this month!
Run the most cyber capable open model at home 🔥
GLM 5.3 EXL3 for 3x DGX Sparks with TensorFold on your 3x DGX Sparks is now LIVE
- Full GLM (not flash)
- Total 5M KV cache
- Yes, ~5,230,000 tokens
- 500k context
- Decode ~31 t/s in prose, 41 t/s in code
- 450k needle passed
There is a catch:
Currently, only 1 stream is supported with 500k context. However, you can run multiple conversation, each on its stream, and the KV cache pools holds them all.
I'm working with @ashxhart to fix it next, so we could actually run concurrent streams with 500k context each.
Excellent quality for coding! Working on a 4x DGX Spark version next. Expect further improvements!
Get it here:
mia-ai.net/models/GLM-5.3-EX…
Yo, @MiaAI_lab — any chance you’d consider building a 3× DGX Spark recipe for the full GLM-5.3?
Ideally something heavily optimized for fast inference/TPS while keeping as much quality as possible. I’m running 3 Sparks now and would love a turnkey TP3 setup similar to your GLM-5.3 Flash work.
Unless, of course, you think there’s a newer/better open model right around the corner that would be more worth targeting 👀
I just bought one of the LAST GB10 rigs (HP ZGX Nano G1n) from ProAdvantage! I will soon have 3 sparks at home, I'm excited to load even better models released over the coming months!
V6 send ✅
This one had a couple moves that felt way harder than they looked — especially that tiny hold in the middle.
Getting to it was one thing.
Having to MATCH on it was another. 😅
Held it together through the crux and finished the send with the match up top.
Definitely one of those climbs where the video does not fully communicate how bad the hold feels.
🧗♂️ V6
✅ Send
#climbing #bouldering #v6 #rockclimbing #climber #boulder #climbingtraining #indoorclimbing
Mia has the best local model recipes, she fine tunes each model to work with different quantities of DGX Sparks. Check out her site, and maybe find your way to the donation part of her site - she deserves every penny!
Come check out my website.
It includes a list of my latest & greatest recipes.
The "Install Model" button gives you a ready-to-use prompt for Claude, Codex, or others. It adapts the recipe to your setup, starts the server and tests it.
mia-ai.net/models
It seems as if "Space Bunny" is likely the new MiniMax model. From everyone's testing, where does it rank? I fully understand it is NOT near Opus 5.5... but how does it rank with GLM 5.3 flash, DeepSeek 4.1 flash, or other Chinese models? @bijanbowen @MiaAI_lab (I trust you both)
I may be able to aquire an R1 (device made by @rabbit_hmi). Anyone have experience with the OS3? Is it useful for your agentic work?
Forrest MacDougall retweeted
Welcome back to Hermes Agent, Claude
New official plugin that uses Claude SDK without the tradeoffs to enable Claude Code subscriptions to work in Hermes Agent again!
Check it out and install it here: hermes-agent.nousresearch.co…
Spent time with Grok 4.7. Verdict: this is 4.6.1, not 4.7 lol
❌ 3D models, the classic social media test, are underwhelming
✅ Research
✅ Code review of other models' work
✅ A few niche use cases
💰 Pricing and speed are the silver lining
Not a flagship. It's a solid second-opinion model.
This V7 just kept going… and going… 😅
Long roof, big moves, terrible places to get pumped, and somehow I held it together.
Definitely one of those climbs where halfway through you start wondering who decided the route needed THAT many moves 😂
Pretty happy with the send though. 💪
#bouldering #climbing #rockclimbing #v7 #boulderinglife #climber #climbingtraining #sendit #roofclimbing
This is incredible!!!
Always wanted to play with my physical Pokémon cards in AR, so I unleashed Astra + CLAD on it… they did not disappoint!
⚡️🥽
Basically a fully playable proto with all 151 mons from SV 151.
@specs @specsfordevs
Wow, @MiaAI_lab does it again! She really is a magician! 🪄
The Need for Speed ⚡️
INSANE DeepSeek v4.1 Flash update on 2x DGX Spark 🔥
+33 FASTER prose on single stream
42 tokens per sec vs old 32
50% FASTER prose on 2 concurrent streams
63.7 tokens per sec vs old 42.5
That puts 2x setup at roughly the same speed as DeepSeek v4.1 Flash running on 3–4x DGX Sparks before the update 🤯
Get it here:
github.com/MiaAI-Lab/DeepSee…
Well said 👏
I’m sorry but America did not spend 250 years turning “what’s over there?” into the most powerful civilization on earth just to finally arrive at the most consequential technological frontier in generations and decide actually we’re scared of frontiers now. We ARE the fucking frontier. We cross oceans. We build railroads across continents. We split atoms. We put men on the moon. We turn sand into computers and computers into a global nervous system.
Of COURSE AI is terrifying. The frontier is SUPPOSED to scare the shit out of you. That’s how you know you have, in fact, located the fucking frontier.
The answer is not recklessness, but it sure as fuck isn’t paralysis either. America’s comparative advantage is an unusually high national tolerance for ambitious weirdos staring at impossible things and saying “Yeah... I think I can build it.” That trait is rarer than oil, harder to manufacture than semiconductors, and responsible for a frankly preposterous amount of the modern world.
And you absolutely will not fucking regulate it out of us.