Antonio Oliveira

1-on-1 mentorship and fractional CTO services.

[Antonio Oliveira @oliveiraacc]

new physics of value:

Wealth = (Taste × Narrative × Distribution) × Time Compression

[Antonio Oliveira @oliveiraacc]

@levelsio They oh-my-pi instead of opencode.
Try adding some custom advisors on into it

[Antonio Oliveira @oliveiraacc]

@no_stp_on_snek Wow, sounds like a pretty productive day off!
I wish all mlx, metal, swift, fused kernels could be easily shared… in rust 🚀

[Antonio Oliveira @oliveiraacc]

@volosatovde Thanks!

That was the first sketches, soon it will be even more consistent.
Programmatically, setting up ltx ic-lora to reference reusable assets, that’s the main bottleneck to scale this into a truly ai content factory!

[Antonio Oliveira @oliveiraacc]

my first attempt to generate content using only local resources with macbook m5 max 128gb

krea, ltx, mlx

[Antonio Oliveira @oliveiraacc]

@ctatedev Is it Zig only? Why not Rust? AI struggles with zig syntax and breaking changes in sdk, requiring a cheat sheet with a good set of hints.

[Antonio Oliveira @oliveiraacc]

my first attempt to generate content using only local resources with macbook m5 max 128gb

krea, ltx, mlx

[Antonio Oliveira @oliveiraacc]

@alexellisuk @UnslothAI @Alibaba_Qwen I added a simple duplicated messages checker in harness. 😉

[Antonio Oliveira @oliveiraacc]

@anemll @deepseek_ai Good work 👏

[Antonio Oliveira @oliveiraacc]

DSpark is up to 400% faster than MTP.

[Daniel Han @danielhanchen]

DeepSeek just released DSpark for V4 Flash & Pro, a new speculative decoding method boosting throughput by 51% to 400%!

DS also showed DSpark works well for other models like Gemma & Qwen

Github: https://t.co/EGVYpc1kcK
Paper: https://t.co/TaBMRVlaW9
HF: https://t.co/289jVU2pxh

[IBM News @IBMNews]

The world’s first sub‑1 nanometer node chip is here.

Delivering 70% greater energy efficiency, this breakthrough powers a new era of computing that’s more capable while using less energy.

Dig into this next-gen tech: https://t.co/NkzAahH49S

[Antonio Oliveira @oliveiraacc]

Wow, that’s really impressive, can’t wait to try this one

[Ornith @ornith_]

Ornith-1.0-9B is easily deployable on edge devices, achieves best-in-class performance among models of similar size (43.1 for Terminal-Bench 2.1 and 69.4 for SWE Verified ), and matches or exceeds the performance of much larger models, including Gemma 4-31B and Qwen 3.6-35B.

Huggingface: https://t.co/GgJ8Upgkbl

[Antonio Oliveira @oliveiraacc]

@gabu3d_pl Whats the prefill speed?

[Antonio Oliveira @oliveiraacc]

That’s so cool!

[Robert Bye @RobertJBye]

Ok so this Teenage Engineering microphone has quickly become my favourite way to talk to Claude, especially in loud offices. Love how tactile it is, and I feel like a radio operator talking to my agents 😂

[Antonio Oliveira @oliveiraacc]

@hardmaru is back 😎

Impressive work

[Sakana AI @SakanaAILabs]

Introducing Sakana Fugu: A full multi-agent orchestration system accessible via a single model API.

Our ‘Fugu Ultra’ model matches the performance of Fable and Mythos, delivering frontier capability without the risk of export controls.

Try it: https://t.co/hhO6qTawgb 🐡

[Guohao Li 🐫 @guohao_li]

just in case you don’t know, the company behind GLM 5.2 is publicly listed, and its stock has gone up 15× in the past six months

[Patrick C Toulme @PatrickToulme]

I ran GLM 5.2 with OpenCode harness against Claude Opus this week deployed locally.

Bottom line: It is a real frontier coding model and insanely good for the price (free). Open source model + open source harness + local serving on my own chips is an amazing value proposition.

Some Notes:

1. Tool calling is very good — spun up nested subagents on its own, multiple levels deep

2. Very good at research and planning including long range plans

3. It built a cell based terminal renderer at near Opus quality. I still lean Claude, but most people couldn't tell the outputs apart

4. Opus wins on oneshotting and reading my intent without me explicitly telling it

5. GLM 5.2 is more than enough intelligence for most F500 work IMO

6. GLM 5.2 is good enough to hill climb RL with and to drive further AI development / next generation GLM model. Progress will be much faster for their RL from here.

7. Running my own endpoint = permanent fast mode

8. It wastes thinking tokens writing code in the reasoning block

At this point I would consider GLM 5.2 a true frontier coding model. Getting to this point in coding quality was the hardest part IMO. They will progress quickly from here in RL.

[Antonio Oliveira @oliveiraacc]

Wonder if they will embed AI, if yes, which model they will choose.

[Rockstar Games @RockstarGames]

Pre-orders for Grand Theft Auto VI will officially begin on June 25 on digital storefronts and at other select retailers.

Check out the official cover art, also available as downloadable artwork at https://t.co/XPwC8URCQ4

[Antonio Oliveira @oliveiraacc]

@DeadByDawn1O1 @jun_song Qwen3.6A3B < GLM5.2
Totally different species of models

[Antonio Oliveira @oliveiraacc]

I tried today, it’s extremely naive model, nothing like those Opus4.5 times.
Opus was more holistic thinking, pro active, persistent, creative, this model is lazy and superficial.

[Ahmad @TheAhmadOsman]

3B model with Opus 4.5 performance

VibeThinker 3B (based on Qwen 2.5)

[Antonio Oliveira @oliveiraacc]

Great tips here! local AI FTW!

[AlexAImaginator @TraffAlex]

🖥️ Best Local LLMs for Consumer GPUs — llama.cpp Guide (June 2026)

What I actually run on consumer hardware right now. Every model below runs via llama.cpp with a simple one-liner — no Docker, no Python env, no cloud.

━━━ 8-16GB VRAM ━━━

🔹 Gemma 4-12B (Google)
• Smartest model in this size class — competes with stuff 2× bigger
• Unsloth's MTP GGUFs: 162 tok/s vs 52 tok/s normal (3× speedup)
• Minimum 8GB VRAM recommended for Q4_K_M quant
• GGUF → https://t.co/VWp818MB3D

🔹 LFM2.5-8B-A1B (LiquidAI)
• Hybrid MoE, only 1B active params — absurdly fast for its size
• Perfect for 8-12GB cards, MacBooks, or anyone on a tight budget
• GGUF → https://t.co/ZbOs4mXJDq

━━━ 16-32GB VRAM ━━━

🔹 Qwen3.6-27B (Qwen)
• Scored 1.00 on tool-efficiency benchmarks — best local agent available
• 40 deterministic tasks, 32k/128k context needle tests — all passed
• GGUF → https://t.co/n7K3sPvliE
• MTP version (faster) → https://t.co/gwdfnJTzcy

🔹 Qwopus3.6-27B-v2 (Jackrong)
• Best quantization of Qwen3.6-27B — topped 5 agent & coding benchmarks (1200 samples)
• If you're running Q4, this is the one to grab
• GGUF → https://t.co/tV1DFqXnOD
• MTP version → https://t.co/PMqz7V5ewv

🔹 Gemma 4-31B QAT (Google/Unsloth)
• QAT variant with MTP draft head: 76-125 tok/s (1.67× speedup)
• Excellent for multi-agent / subagent workflows
• GGUF → https://t.co/FgVsUX0YOB

🔹 Nex-N2-Mini (Nex AGI)
• Post-train of Qwen3.5-35B-A3B — MoE with only 3B active params
• Fits on 16GB+ VRAM, overflow loads from system RAM
• Adaptive thinking saves ~20% tokens with no quality loss
• For deep multi-step reasoning, nothing in this size comes close
• GGUF → https://t.co/oyC522a8Eh

━━━ Quick Picks ━━━

• 16GB all-rounder → Gemma 4-12B with MTP GGUFs
• 32GB all-rounder → Qwen3.6-27B / Qwopus-v2
• Agents & tool use → Qwen3.6-27B or Qwopus Q4
• Deep reasoning → Nex-N2-Mini (MoE, fits 16GB+)
• Tight budget → LFM2.5-8B-A1B
• Cheapest full build: 1× used RTX 3090 (24GB) + rest of PC ≈ $1000-1500

━━━ Setup on Windows ━━━

1. Download llama.cpp → https://t.co/et0J7Swua7 (latest .zip)
2. Extract to any folder (e.g. C:\llama.cpp)
3. Download a .gguf from the links above (Q4_K_M or Q5_K_M for best quality/speed balance)
4. Run one of the commands below depending on your hardware

━━━ Launch Commands ━━━

SINGLE GPU — Standard model (no MTP):

llama-server.exe ^
-m C:\models\Qwen3.6-27B-Q5_K_M.gguf ^
--ctx-size 180000 ^
--flash-attn on ^
--cache-type-k q4_0 ^
--cache-type-v q4_0 ^
--batch-size 1024 --ubatch-size 512 ^
-ngl 100 ^
-np 1 ^
--port 8080 ^
--jinja

SINGLE GPU — MTP model (faster inference):

llama-server.exe ^
-m C:\models\Qwen3.6-27B-MTP-Q5_K_M.gguf ^
--ctx-size 180000 ^
--flash-attn on ^
--cache-type-k q4_0 ^
--cache-type-v q4_0 ^
--batch-size 1024 --ubatch-size 512 ^
--spec-type draft-mtp ^
--spec-draft-n-max 3 ^
-ngl 100 ^
-np 1 ^
--port 8080 ^
--jinja

DUAL GPU — Split across two cards:

llama-server.exe ^
-m C:\models\Qwen3.6-27B-Q5_K_M.gguf ^
--ctx-size 180000 ^
--flash-attn on ^
--cache-type-k q4_0 ^
--cache-type-v q4_0 ^
--batch-size 1024 --ubatch-size 512 ^
-ngl 100 ^
--tensor-split 0.55,0.45 ^
--main-gpu 0 ^
-np 1 ^
--port 8080 ^
--jinja

DUAL GPU + MTP + Vision (multimodal):

llama-server.exe ^
-m C:\models\Qwen3.6-27B-MTP-Q5_K_M.gguf ^
--ctx-size 180000 ^
--flash-attn on ^
--cache-type-k q4_0 ^
--cache-type-v q4_0 ^
--batch-size 1024 --ubatch-size 512 ^
--spec-type draft-mtp ^
--spec-draft-n-max 3 ^
-ngl 100 ^
--tensor-split 0.60,0.40 ^
--main-gpu 0 ^
-np 1 ^
--port 8080 ^
--jinja ^
--mmproj C:\models\mmproj-F16.gguf

━━━ Parameter Breakdown ━━━

-m <path>
Path to your .gguf model file. Change this to wherever you downloaded it.

--ctx-size 180000
Context window in tokens. 180k = huge context for long conversations or big codebases.
Reduce to 32768 or 65536 if you don't need long context — uses less VRAM.

--flash-attn on
Flash Attention — dramatically speeds up inference and reduces VRAM usage.
Works on RTX 30xx/40xx/50xx. Always enable this.

--cache-type-k q4_0 / --cache-type-v q4_0
Quantizes the KV cache (key/value attention cache) to 4-bit.
This is what makes 180k context fit in VRAM. Without it, huge contexts eat all your memory.
Quality impact is minimal — this is a free performance win.

--batch-size 1024 / --ubatch-size 512
batch-size = how many tokens are processed in one forward pass (throughput).
ubatch-size = micro-batch actually sent to the GPU per step.
Higher = faster prompt processing but needs more VRAM.
If you run out of VRAM, lower these (e.g. 512/256).

-ngl 100
Number of layers to offload to GPU. 100 = all layers on GPU (full offload).
This is what you want if the model fits in your VRAM.
If it doesn't fit, reduce this (e.g. -ngl 40) — remaining layers run on CPU/RAM.

--tensor-split 0.55,0.45
How to split model layers across multiple GPUs. Values are ratios.
0.55,0.45 = GPU 0 gets 55% of layers, GPU 1 gets 45%.
Adjust based on your VRAM — give more to the card with more memory.
Example: 0.70,0.30 for a 24GB + 12GB setup.
Not needed for single GPU setups.

--main-gpu 0
Which GPU handles the batch computation (the "orchestrator").
Set to 0 (your primary GPU). The other GPU(s) handle their assigned layers.
Minor performance impact — usually just leave it at 0.

-np 1
Number of parallel slots (concurrent requests). 1 = one user at a time.
Increase to 2-4 if you want multiple clients connected simultaneously.
Each extra slot uses additional VRAM for its own KV cache.

--port 8080
Which port the server listens on. Change if port 8080 is busy.

--jinja
Enables Jinja2 template processing — required for proper chat formatting.
Most modern models expect this. Always include it.

--spec-type draft-mtp
Enables Multi-Token Prediction (MTP) speculative decoding.
Only works with MTP GGUF models (downloaded separately).
The model predicts multiple tokens at once and verifies them — big speed boost.

--spec-draft-n-max 3
How many tokens the MTP draft head proposes per step.
3 is a good default. Higher = potentially faster but more VRAM and may reduce quality.

--mmproj <path>
Path to the multimodal projector file (for vision models).
Enables image understanding — paste screenshots into the web chat.
Only needed if you want vision capabilities. Omit for text-only use.

━━━ Your Hardware → Your Command ━━━

Single GPU (8-24GB VRAM):
Use the "Single GPU" command. Change -m to your model path.
8GB card → Gemma 4-12B Q4 or LFM2.5-8B
12GB card → Gemma 4-12B Q5/Q6
16GB card → Gemma 4-31B QAT Q4 or Nex-N2-Mini
24GB card → Qwen3.6-27B Q4/Q5, Qwopus-v2, Gemma 4-31B QAT Q5/Q6

Dual GPU:
Use the "Dual GPU" command. Adjust --tensor-split based on your VRAM ratio.
24GB + 24GB → --tensor-split 0.50,0.50
24GB + 12GB → --tensor-split 0.70,0.30
24GB + 8GB → --tensor-split 0.75,0.25

Want speed? Use MTP versions of models with the "MTP" commands.

Want vision? Add --mmproj with the projector file from the model's HuggingFace repo.

5. Once running, you get:
• Web chat UI → http://localhost:8080
• OpenAI-compatible API → http://localhost:8080/v1
• Playground → http://localhost:8080/playground

━━━ Why /v1 API Is the Killer Feature ━━━

One local endpoint replaces your entire cloud API bill. The /v1 endpoint is drop-in OpenAI-spec compatible — every tool that speaks OpenAI just works. No custom code, no glue layer.

Works out of the box with:
• IDEs: Cursor, Continue, Windsurf, Cline, Roo Code
• CLI tools: aider, Open Interpreter, OpenCode
• Frameworks: LangChain, LlamaIndex, LiteLLM
• Any OpenAI SDK (Python, Node, Go, Rust)

Why this beats cloud APIs:
• 100% private — code never leaves your machine
• $0 per token — no rate limits, no quotas, no surprise bills
• Works fully offline
• Zero telemetry, no training on your data
• Swap models by dropping in a different .gguf — no app changes needed
• Run 32k–128k context windows without burning money

Good combos:
• Cursor + Qwopus-v2 → near-frontier quality, zero API cost
• Continue + Qwen3.6-27B → best local coding agent
• aider + Gemma 4-12B MTP → 162 tok/s, feels instant
• OpenCode + Nex-N2-Mini → deep reasoning on 16GB

Set any OpenAI-compatible client to your local endpoint:
set OPENAI_API_KEY=sk-dummy (any non-empty string works)
set OPENAI_BASE_URL=http://localhost:8080/v1
# every OpenAI-compatible tool now hits your local GPU

Shoutouts: @0xSero @rS_alonewolf @witcheer @UnslothAI @LottoLabs