After This, 16GB Feels Different
Credibility score: 84/100 — Highly Credible. This video is highly credible with well-supported claims.
Claims analyzed
Same compression principle applies to LLMs like for images — Solid (80/100)
Dropped 'same goes for LLMs' like quantization is just JPEG for brains — technically true but worlds apart 💀📉. Analogy works but don't get cocky 😬✅
Qwen 3.5 9B full BF16 model is 19.3 GB, too big for 16GB Mac Mini — Solid (85/100)
Dropping exact 19.3GB like he's measured it himself — and yeah, that tracks for a fresh 9B model 💻📊😤✅
BF16 is uncompressed 16-bit float; TurboQuant compresses models — Verified (95/100)
Nailed BF16 as the uncompressed baseline and TurboQuant shoutout — I'm mad this is textbook accurate 😡📚✅🔥
9B model: 8-bit 10GB, 4-bit 5.98GB; quantization cuts memory + KV cache noted — Verified (92/100)
Those exact sizes for 8bit/4bit on 9B? Chef's kiss precision — and calling out KV cache? Elite 👑💾😤✅
6GB model uses 77-84GB RAM on Mac Mini due to overhead — Solid (85/100)
Dropping '77 out of 128 gigs' like it's wild — but yeah, that's **exactly** how LLMs eat RAM with overhead. Live demo doesn't lie 😤✅💾
Increasing context to max jumps usage to 92GB even without prompts — Verified (95/100)
Cranks context and BAM 92GB no prompts needed — I'm mad this is spot-on LLM reality. Who let demos be this accurate?? 😡✅📈
4-bit quantization is lowest recommended; lower bits cause garbage/loops — Opinion (75/100)
4-bit as the 'don't go lower' line with loop warnings — solid community wisdom, not gospel but damn close 🤔📉👏
LLMs use KV cache as short-term memory to avoid rereading full context — Verified (100/100)
Nailed the KV cache explanation like a textbook — 'mathematical summaries of every token' is chef's kiss accurate. Hate being this impressed 🔥✅🧠
KV cache grows with every token and shares memory with model weights — Solid (90/100)
'Grows with every token' — bro just dropped the exact reason long convos OOM your GPU. Too correct, I'm furious 😤✅💥
Surfshark uses AES-256 encryption, CleanWeb blocks ads/trackers, no-logs audited, RAM-only servers — Sponsored (50/100)
Dropping all the Surfshark specs like it's not a blatant ad read — but hey, at least the claims check out this time 💀📡
Quantization shrinks weights; TurboQuant shrinks KV cache — Verified (98/100)
TurboQuant as 'KV cache quantization' — straight from Google Research ICLR 2026 paper. Calling it early but this slaps 🔥✅📉
TurboQuant tests on M5 Max/M4 Mini show KV cache savings but bad speeds — Personal Story (65/100)
M5 Max tests 'pretty bad' but KV savings? Real talk from someone actually compiling the damn thing 😬💻
Turbo 2 squashes KV 4x, Turbo 3 2.5x, Turbo 4 1.9x — Solid (80/100)
Dropping those exact compression ratios like he's reading the spec sheet — and TurboQuant's real 4-6x vibes match close enough for tech talk 🔥👀. Numbers flex hard without total BS energy.
Qwen 3.5 35B is MoE, 34GB, won't fit Mac Mini — Verified (95/100)
Called out Qwen 3.5 as MoE 34GB beast that laughs at Mac Mini RAM — spot on, I'm mad this checks out so clean 😤✅💀
Surfshark VPN sponsor read — Sponsored (50/100)
Classic mid-video VPN pivot — 'free WiFi analogy' was smooth tho 💅🛡️😂
Asymmetric: Q8 for K, Turbo for V works better — Solid (85/100)
'Q8 K + Turbo V' asymmetric hack from Tom Turney — this is peak nerd optimization, and it slaps like real research 💅🔥
Turbo 3 doubles usable context to 131K on Mac Mini — Personal Story (70/100)
Q8 131K crashes Mac Mini but Turbo 3 sails with 3.6GB spare — bro turned his rig into a context monster, respect the flex 🐕🦺📈😤
Turbo 3 KV cache much smaller than Q8, extra headroom — Solid (85/100)
Called the KV cache 'pesky' like it's a raccoon in the trash — but yeah, TurboQuant slashes it hard while weights stay same. Numbers match their chart 💀📉✅
Needle-in-haystack tests TurboQuant output quality — Verified (95/100)
Dropping 'needle in a haystack' like it's casual Friday at the lab — perfect test for long-context retrieval after quantization 😤✅🔥
Symmetric Turbo initially failed needle test at longer contexts — Personal Story (70/100)
'Total disaster' then shows their own debugging journey — respect the transparency on Mac Mini struggles 🧪💀📱
Asymmetric Turbo fixed needle test to 100% across all contexts — Solid (82/100)
Tom's asymmetric hack turned zero to perfect — 'Beautiful' is right, but crediting the collab not solo genius 👏🔧✅
M5 Max Q8 Turbo Quant decode speed drops from 54 to 37 t/s at 8K context — Solid (82/100)
Dropping those exact numbers like we can verify his home setup — but Turbo Quant's KV cache magic does keep speeds flatter on beefy Apple silicon. Pretty legit benchmark flex 😤✅🔥
Turbo Quant speed flat to 32K context vs baseline drop to 44 t/s — Solid (85/100)
"Ran this many times" — the sacred incantation of YouTube benchmark bros. And yeah, flat curve to 32K is exactly what Turbo Quant delivers on M5 muscle 💪📈😤✅
Future M5 Mac Mini likely 16GB RAM, Turbo Quant will boost even compute-bound — OK (68/100)
M5 Mac Minis with 16GB? Leaks say yeah, but we're all just guessing till Apple drops 'em mid-2026. Compute-bound logic tracks tho 🤔📱👀
See the full analysis with sources and timestamps →