ingest(raw+wiki): Paul Couvert — best local models for laptop (cross-platform companion to MLX MoE)

2026-06-30 17:12:25 +02:00 · 2026-06-30 17:12:25 +02:00 · 8c4cdbf1a0
commit 8c4cdbf1a0
parent f4aca5076c
5 changed files with 303 additions and 1 deletions
--- a/raw/xpost/2026-06-29_paulcouvert-local-llm-laptop-guide.md
+++ b/raw/xpost/2026-06-29_paulcouvert-local-llm-laptop-guide.md
@ -0,0 +1,91 @@
+---
+type: xpost
+source_url: https://x.com/itsPaulAi/status/2071717503105900993
+author: "@itsPaulAi"
+date: 2026-06-29
+posted_by: Pit Weber (@PWeber)
+topic: "Tips & Tricks (Topic 27)"
+tags: [xpost, local-llm, laptop, qwen, gemma, parakeet, unsloth, llama-cpp, lm-studio, quantization, voice-to-text, diffusion, edge-ai, google-ai-edge-gallery]
+---
+
+# Paul Couvert: Best Local Models You Can Run on a Laptop Today
+
+**Source:** [X-Post by @itsPaulAi](https://x.com/itsPaulAi/status/2071717503105900993)
+**Posted:** 2026-06-29
+**Shared by:** Pit Weber in OME-Gruppe, Topic "Tips & Tricks" (Topic 27)
+
+## Post Content
+
+Paul Couvert (@itsPaulAi) lists the best local models for laptop-class hardware — explicitly noting you don't need a $10k Mac Studio:
+
+> The best local models you can run on a laptop today (without a $10k Mac Studio):
+>
+> 1. **Qwen3.6-27B** — by far the best for coding agents (or **Qwen3.6-35B-A3B** for a faster option)
+> 2. **Gemma 4 12B** — everyday tasks and questions
+> 3. **Parakeet 0.6B v3** — basically the best voice-to-text model you can find
+> 4. **Gemma 4 E4B** — surprisingly good for the size, can run offline on a phone
+> 5. **Gemma 4 26B diffusion** — the most tok/s you can have with a local model
+>
+> Tips:
+> - Use **quantized versions from Unsloth** when running on a laptop — precision/size/speed are fantastic
+> - **LM Studio** or raw **llama.cpp** are the best runtime options
+> - **Google AI Edge Gallery** on iOS/Android
+
+## Model Overview Table
+
+| # | Model | Size | Use Case | Key Strength |
+|---|-------|------|----------|--------------|
+| 1 | Qwen3.6-27B | 27B | Coding agents | "By far the best for coding" |
+| 1b | Qwen3.6-35B-A3B | 35B (MoE, 3B active) | Coding (faster alt) | MoE sparse activation for speed |
+| 2 | Gemma 4 12B | 12B | Everyday tasks/questions | General-purpose balanced |
+| 3 | Parakeet 0.6B v3 | 0.6B | Voice-to-text (STT) | "Best voice-to-text model you can find" |
+| 4 | Gemma 4 E4B | ~4B | Edge / phone | Runs offline on phone |
+| 5 | Gemma 4 26B diffusion | 26B | Speed-focused | "Most tok/s for a local model" |
+
+## Practical Tips
+
+| Tip | Detail |
+|-----|--------|
+| **Quantization** | Use Unsloth quantized versions — "precision/size/speed are fantastic" |
+| **Runtime** | LM Studio or raw llama.cpp recommended |
+| **Mobile** | Google AI Edge Gallery app (iOS/Android) for on-device inference |
+
+## Engagement
+
+- 138 likes
+- 12 reposts
+- 18 replies
+- 170 bookmarks
+- ~13.5K views
+
+## Context: Complements Jun Song MLX MoE Post
+
+This post complements the [Jun Song MLX MoE post](https://x.com/jun_song/status/2071625328083227118) (wikified as [[concepts/llm/mlx-moe-local-ai-optimization.md]]):
+
+| Dimension | Jun Song | Paul Couvert |
+|-----------|----------|-------------|
+| **Focus** | Apple Silicon / MLX optimization | Any laptop (cross-platform) |
+| **Key insight** | MoE > Dense on MLX | Quantized models from Unsloth for any laptop |
+| **Top coding pick** | Minimax-M3.0 (dq), DeepSeek-v4-Flash (dq) | Qwen3.6-27B (or 35B-A3B MoE variant) |
+| **Dense 27B on Mac** | "Painfully slow" — avoid | Recommends Qwen3.6-27B for coding (on any laptop, quantized) |
+| **Resolution** | Different deployment contexts: Jun Song = Mac/MLX native; Paul = any laptop with Unsloth quants | |
+| **Common ground** | Both recommend Qwen3.6-35B-A3B (MoE variant) — Jun Song implicitly (MoE preference), Paul explicitly (faster option) | |
+
+**Key distinction:** Jun Song warns about dense 27B on Mac/MLX specifically. Paul recommends Qwen3.6-27B for coding agents on any laptop — likely with Unsloth quantization, which changes the performance profile. The MoE variant (Qwen3.6-35B-A3B) is where both converge: Jun Song's MoE preference + Paul's "faster option."
+
+## Relevance to Our Setup
+
+1. **Qwen3.6-35B-A3B MoE variant** — aligns with Jun Song's MoE recommendation and our model routing (MoE models as fallbacks)
+2. **Unsloth quantization** — practical tip for running models on commodity hardware, complements Cloud-Exit thesis
+3. **Parakeet 0.6B for STT** — relevant for OpenClaw voice interaction, potential local Whisper alternative
+4. **Gemma 4 E4B on phone** — edge inference extends Cloud-Exit to mobile, connects to OpenClaw mobile apps
+5. **LM Studio / llama.cpp** — runtime options already mentioned in our wiki (GLM-5.2 self-hosting via LM Studio)
+
+## Related Wiki Pages
+
+- `wiki/concepts/llm/mlx-moe-local-ai-optimization.md` — Jun Song's MLX MoE post (direct complement)
+- `wiki/concepts/hardware/cloud-exit-and-local-superiority.md` — Cloud-Exit thesis
+- `wiki/concepts/llm/llm-model-catalog.md` — Consolidated model catalog
+- `wiki/concepts/llm/chinese-model-cost-routing.md` — DeRonin's cost routing (same Qwen models)
+- `wiki/concepts/hardware/edge-inference-als-cloud-alternative.md` — Edge inference (AMD Strix Halo)
+- `wiki/tools/hermes-desktop.md` — Hermes MoA (related ensemble pattern)
--- a/wiki/concepts/llm/local-llm-laptop-guide.md
+++ b/wiki/concepts/llm/local-llm-laptop-guide.md
@ -0,0 +1,195 @@
+---
+created: 2026-06-29
+updated: 2026-06-29
+sources:
+  - xpost/2026-06-29_paulcouvert-local-llm-laptop-guide.md
+  - concepts/llm/mlx-moe-local-ai-optimization.md
+  - concepts/hardware/cloud-exit-and-local-superiority.md
+  - concepts/llm/llm-model-catalog.md
+  - concepts/llm/chinese-model-cost-routing.md
+tags: [concept, llm, local-ai, laptop, quantization, unsloth, qwen, gemma, parakeet, llama-cpp, lm-studio, edge-ai, voice-to-text, diffusion, moe]
+---
+
+# Local LLM Laptop Guide — Best Models Without a $10k Mac Studio
+
+> **TL;DR:** Paul Couvert (@itsPaulAi) lists the best local models for laptop-class hardware. Top picks: Qwen3.6-27B for coding, Gemma 4 12B for everyday tasks, Parakeet 0.6B v3 for voice-to-text, Gemma 4 E4B for phone-class edge inference, and Gemma 4 26B diffusion for maximum tok/s. Key tips: use Unsloth quantized versions, run via LM Studio or llama.cpp, try Google AI Edge Gallery on mobile. Complements Jun Song's MLX MoE optimization post — both converge on the Qwen3.6-35B-A3B MoE variant.
+
+## Source
+
+| Source | Author | Date | Engagement |
+|--------|--------|------|------------|
+| [X-Post: "Best local models on a laptop"](https://x.com/itsPaulAi/status/2071717503105900993) | Paul Couvert (@itsPaulAi) | 2026-06-29 | 138 likes, 12 reposts, 170 bookmarks, ~13.5K views |
+
+**Shared by:** Pit Weber in OME-Gruppe, Topic "Tips & Tricks" (Topic 27)
+
+---
+
+## The Model Stack
+
+### 1. Coding: Qwen3.6-27B (or Qwen3.6-35B-A3B for speed)
+
+| Dimension | Qwen3.6-27B (Dense) | Qwen3.6-35B-A3B (MoE) |
+|-----------|---------------------|------------------------|
+| **Total params** | 27B | 35B |
+| **Active params/token** | 27B (all active) | 3B (sparse) |
+| **Best for** | Coding agents — "by far the best" | Coding, but faster |
+| **Laptop fit** | Needs quantization (Unsloth) | MoE sparse activation = faster on limited hardware |
+| **Cross-ref** | Jun Song warns: slow on Mac/MLX | Jun Song's MoE preference aligns |
+
+> **⚠️ Tension with Jun Song:** Jun Song explicitly says dense 27B on Mac is "painfully slow." Paul recommends it for coding on any laptop — likely assuming Unsloth quantization on non-Mac hardware. The **MoE variant (35B-A3B)** is where both agree: it's the faster, better-suited option for constrained hardware. See [resolution below](#resolving-the-dense-27b-tension).
+
+### 2. Everyday Tasks: Gemma 4 12B
+
+| Dimension | Detail |
+|-----------|--------|
+| **Size** | 12B |
+| **Best for** | Everyday tasks and questions |
+| **Why** | Balanced size-to-quality ratio for general use |
+| **Runtime** | LM Studio, llama.cpp, or Google AI Edge Gallery |
+
+### 3. Voice-to-Text: Parakeet 0.6B v3
+
+| Dimension | Detail |
+|-----------|--------|
+| **Size** | 0.6B — tiny |
+| **Best for** | Speech-to-text (STT) |
+| **Claim** | "Basically the best voice-to-text model you can find" |
+| **Relevance** | Local Whisper alternative; relevant for OpenClaw voice interaction pipelines |
+| **Edge fit** | Small enough to run on any laptop or phone |
+
+### 4. Edge / Phone: Gemma 4 E4B
+
+| Dimension | Detail |
+|-----------|--------|
+| **Size** | ~4B |
+| **Best for** | On-device inference, offline phone use |
+| **Claim** | "Surprisingly good for the size" |
+| **Mobile deployment** | Google AI Edge Gallery (iOS/Android) |
+| **Cross-ref** | Extends [Cloud-Exit thesis](../hardware/cloud-exit-and-local-superiority.md) to mobile — connects to [OpenClaw mobile apps](../../tools/openclaw.md) |
+
+### 5. Speed Champion: Gemma 4 26B Diffusion
+
+| Dimension | Detail |
+|-----------|--------|
+| **Size** | 26B |
+| **Type** | Diffusion-based LLM |
+| **Best for** | Maximum tok/s on local hardware |
+| **Claim** | "The most tok/s you can have with a local model" |
+| **Architecture note** | Diffusion LLMs generate tokens via iterative denoising — different paradigm from standard AR models. See [[post-transformer-llm-architectures.md]] for DeepMind's diffusion LLM research (10× efficiency gains). |
+
+---
+
+## Practical Deployment Tips
+
+### Quantization: Unsloth
+
+[Unsloth](https://github.com/unslothai/unsloth) provides quantized model versions optimized for laptop deployment:
+
+| Benefit | Detail |
+|---------|--------|
+| **Precision/size/speed** | "Fantastic" balance — Paul's explicit recommendation |
+| **Compatibility** | Works with LM Studio, llama.cpp |
+| **How to use** | Download Unsloth quantized GGUF files → load in LM Studio or llama.cpp |
+
+### Runtime Options
+
+| Runtime | Type | Best For |
+|---------|------|----------|
+| [LM Studio](https://lmstudio.ai) | GUI app | Easy setup, model browsing, chat interface |
+| [llama.cpp](https://github.com/ggerganov/llama.cpp) | CLI / library | Maximum control, minimal overhead, embedding in apps |
+| [Google AI Edge Gallery](https://ai.google.dev/edge) | Mobile app | On-device inference on iOS/Android |
+
+---
+
+## Resolving the Dense 27B Tension
+
+Paul recommends Qwen3.6-27B (dense) for coding. Jun Song says dense 27B on Mac is "painfully slow." This isn't a contradiction — it's a **context distinction**:
+
+| Factor | Paul's Context | Jun Song's Context |
+|--------|---------------|-------------------|
+| **Hardware** | Any laptop (x86, ARM, Mac) | Specifically Apple Silicon / MLX |
+| **Quantization** | Unsloth quantized (explicitly recommended) | Dequantized (dq) MoE models |
+| **Framework** | llama.cpp / LM Studio (CPU/GPU offload) | MLX (Apple native) |
+| **Dense 27B verdict** | ✅ Good with Unsloth quants | ❌ "Painfully slow" on MLX |
+| **MoE 35B-A3B verdict** | ✅ "Faster option" | ✅ Implied by MoE preference |
+
+**Resolution:** The performance of dense 27B depends heavily on **framework + quantization**:
+- **Unsloth quantized on llama.cpp** (Paul's path): tractable on commodity laptops
+- **Unquantized/dq on MLX** (Jun Song's path): slow because MLX doesn't optimize for dense
+- **MoE variant (35B-A3B)**: optimal on both paths — sparse activation reduces compute per token
+
+**Recommendation:** If you're on a Mac, follow Jun Song's advice (MoE models via MLX). If you're on any laptop with llama.cpp + Unsloth quants, Paul's dense 27B recommendation works. When in doubt, pick the MoE variant (Qwen3.6-35B-A3B) — it's the convergence point.
+
+---
+
+## Complement to MLX MoE Optimization
+
+This page is the **cross-platform companion** to [[mlx-moe-local-ai-optimization.md]]:
+
+| Dimension | [MLX MoE Optimization](mlx-moe-local-ai-optimization.md) | This Page (Laptop Guide) |
+|-----------|----------------------------------------------------------|--------------------------|
+| **Scope** | Apple Silicon / MLX | Any laptop (x86, ARM, Mac) |
+| **Key insight** | MoE > Dense on MLX (unified memory) | Quantized models work on any laptop |
+| **Model sources** | Minimax-M3.0, DeepSeek-v4-Flash (Chinese MoE) | Qwen3.6, Gemma 4, Parakeet (broader mix) |
+| **Focus** | Why MoE works on Apple Silicon | What to run + how (runtime, quantization) |
+| **Unique coverage** | Info gap thesis (frontier AIs can't tell you) | STT (Parakeet), phone-edge (Gemma E4B), diffusion speed (Gemma 26B) |
+
+**Together:** Jun Song tells you *why* MoE on Mac is better. Paul tells you *what* to run on any laptop and *how*. The MoE variant Qwen3.6-35B-A3B is the overlap — both recommend it, for different reasons.
+
+---
+
+## Model Landscape Summary
+
+| Use Case | Model | Size | Architecture | Hardware Floor |
+|----------|-------|------|--------------|----------------|
+| Coding (best) | Qwen3.6-27B | 27B | Dense | Laptop w/ quantization |
+| Coding (fast) | Qwen3.6-35B-A3B | 35B/3B active | MoE | Laptop |
+| Everyday | Gemma 4 12B | 12B | Dense | Laptop |
+| Voice-to-text | Parakeet 0.6B v3 | 0.6B | STT | Any device |
+| Phone/edge | Gemma 4 E4B | ~4B | Dense | Phone |
+| Speed champion | Gemma 4 26B diffusion | 26B | Diffusion LLM | Laptop (GPU helpful) |
+
+---
+
+## Connection to Our Setup
+
+### Model Routing Alignment
+
+| Model | Paul's Role | Our Role | Match |
+|-------|-------------|----------|-------|
+| Qwen3.6-35B-A3B | "Faster coding option" (MoE) | Not yet in active routing — candidate | Potential addition |
+| Qwen3.6-27B | "Best for coding agents" | Not in active routing | Reference point |
+| Gemma 4 12B | "Everyday tasks" | Not in active routing | Consider for lightweight tasks |
+
+### Cloud-Exit Extension
+
+Paul's guide extends the Cloud-Exit thesis to **laptop-class hardware** — you don't need a Mac Studio, DGX Station, or Strix Halo:
+
+| Hardware Tier | Cloud-Exit Option | Source |
+|---------------|-------------------|--------|
+| Phone | Gemma 4 E4B via Google AI Edge Gallery | Paul Couvert |
+| Laptop (any) | Qwen3.6/Gemma 4 with Unsloth quants via LM Studio/llama.cpp | Paul Couvert |
+| Mac (Apple Silicon) | MoE models (Minimax-M3.0, DeepSeek-v4-Flash) via MLX | Jun Song |
+| Prosumer desktop | AMD Strix Halo (128 GB) | [[../hardware/edge-inference-als-cloud-alternative.md]] |
+| Enterprise desktop | NVIDIA DGX Station (748 GB) | [[../hardware/nvidia-dgx-station-748gb.md]] |
+
+---
+
+## Cross-References
+
+- [[mlx-moe-local-ai-optimization.md]] — Jun Song's MLX MoE post (direct complement, Apple Silicon focus)
+- [[../hardware/cloud-exit-and-local-superiority.md]] — Cloud-Exit thesis (this page extends to laptop tier)
+- [[../hardware/edge-inference-als-cloud-alternative.md]] — AMD Strix Halo (edge inference, x86)
+- [[../hardware/nvidia-dgx-station-748gb.md]] — NVIDIA DGX Station (enterprise tier)
+- [[llm-model-catalog.md]] — Consolidated model catalog
+- [[chinese-model-cost-routing.md]] — DeRonin's cost routing (Qwen models in cloud→local swap)
+- [[post-transformer-llm-architectures.md]] — Diffusion LLMs (Gemma 4 26B diffusion context)
+- [[../../tools/openclaw.md]] — OpenClaw mobile apps (Gemma E4B on-phone connection)
+
+## External Links
+
+- [Original X-Post by Paul Couvert](https://x.com/itsPaulAi/status/2071717503105900993)
+- [Unsloth GitHub](https://github.com/unslothai/unsloth)
+- [LM Studio](https://lmstudio.ai)
+- [llama.cpp](https://github.com/ggerganov/llama.cpp)
+- [Google AI Edge Gallery](https://ai.google.dev/edge)
--- a/wiki/concepts/llm/mlx-moe-local-ai-optimization.md
+++ b/wiki/concepts/llm/mlx-moe-local-ai-optimization.md
@ -123,6 +123,7 @@ This page directly extends the [Cloud-Exit thesis](../hardware/cloud-exit-and-lo
 - [[ai-intelligence-commoditization-thesis.md]] — TheProphet's macro thesis (same models, geopolitics angle)
 - [[llm-model-catalog.md]] — Consolidated model catalog with local deployment options
 - [[../../architecture/model-routing.md]] — Our routing architecture
+- [[local-llm-laptop-guide.md]] — Paul Couvert's cross-platform laptop guide (companion page, any-laptop perspective)
 - [[../../tools/hermes-desktop.md]] — Hermes MoA (Mixture of Agents, related ensemble pattern)

 ## External Links
--- a/wiki/index.md
+++ b/wiki/index.md
@ -2,7 +2,7 @@

 *Auto-generated: 2026-06-23*

- *Letzte Aktualisierung: 2026-06-30 (46. Update — OpenClaw veröffentlicht native iOS- und Android-Apps (29.06.2026). Erste Open-Source-KI-Agent-Companion-App auf beiden Plattformen. Gateway-Pairing via QR/Setup-Code, Voice-Interaktion, Action Approvals, Device Capabilities. ClawHub-Security: 5 malicious skills bypassed screening (Unit 42), 800+ gesamt. Neue Mobile-Apps-Sektion in `tools/openclaw.md`.)*
+ *Letzte Aktualisierung: 2026-06-30 (47. Update — Paul Couvert (@itsPaulAi): Best local models for laptop-class hardware. Qwen3.6-27B/35B-A3B for coding, Gemma 4 12B for everyday, Parakeet 0.6B v3 for STT, Gemma 4 E4B for phone, Gemma 4 26B diffusion for speed. Unsloth quants + LM Studio/llama.cpp. Cross-platform companion to Jun Song's MLX MoE post. Neue Seite `concepts/llm/local-llm-laptop-guide.md`.)*

 ## Architecture

@ -67,6 +67,7 @@
 | [ReDS — Resilient Decentralized Swarm](concepts/llm/reds-resilient-decentralized-swarm.md) | Brian Roemmele's Konzept für dezentrale KI-Entwicklung als Gegenmodell zu Amodei's Zentralismus. Resilient (zensur-resistent) + Decentralized (keine zentrale Kontrolle) + Swarm (koordinierte Akteure). Erweitert Cloud-Exit/Dezentrale-KI-These um politische/soziale Dimension | xpost/2026-06-28_roemmele-reds-decentralized-ai-vs-amodei.md |
 | [Chinese Model Cost Routing — The 87% Cost-Cut Playbook](concepts/llm/chinese-model-cost-routing.md) | DeRonin's 30-day field report: 6 Western→Chinese model swaps, 87% cost reduction, 4% quality drop, revenue unchanged. Swap matrix (Opus→Kimi K2.7, GPT-5.5→Qwen 3.7 Max, Sonnet→GLM 5.2, GPT-mini→MiMo V2.5, GPT-Image→Wan 2.5, Sora→Kling 3.0). Validates Barbell Routing + Factory-for-Gods thesis. Cross-refs to model-router skill | xpost/2026-06-29_deronin-chinese-ai-stack-cost-savings.md |
 | [MLX MoE Optimization — Why Apple Silicon Loves Mixture-of-Experts](concepts/llm/mlx-moe-local-ai-optimization.md) | Jun Song: MLX optimized for MoE not dense. Qwen3.6 27B on Mac = slow/hot. Minimax-M3.0 (dq) + DeepSeek-v4-Flash (dq) = smooth. Info-Gap: frontier AIs can't know this. Triple validation with DeRonin (cost) + TheProphet (macro) for same models | xpost/2026-06-29_junsong-mlx-moe-local-ai-info-gap.md |
+| [Local LLM Laptop Guide — Best Models Without a $10k Mac Studio](concepts/llm/local-llm-laptop-guide.md) | Paul Couvert: Top 5 local models for any laptop. Qwen3.6-27B/35B-A3B (coding), Gemma 4 12B (everyday), Parakeet 0.6B v3 (STT), Gemma 4 E4B (phone), Gemma 4 26B diffusion (speed). Unsloth quants + LM Studio/llama.cpp. Cross-platform companion to MLX MoE page. Resolves dense-27B tension (framework+quantization context) | xpost/2026-06-29_paulcouvert-local-llm-laptop-guide.md |

 ### AGI
 | Seite | Beschreibung | Quellen |
@ -248,3 +249,4 @@
 | `raw/xpost/2026-06-29_deronin-chinese-ai-stack-cost-savings.md` | xpost | DeRonin: "My entire AI stack is now Chinese 🇨🇳" — 87% Cost-Cut Field Report |
 | `raw/blog/2026-06-29_heise-glm52-hacking-cybersecurity.md` | news | Chinas GLM-5.2 erreicht Anthropics Mythos bei der Schwachstellensuche (heise) |
 | `raw/xpost/2026-06-29_junsong-mlx-moe-local-ai-info-gap.md` | xpost | Jun Song: MLX optimized for MoE not dense — local AI info gap on Apple Silicon |
+| `raw/xpost/2026-06-29_paulcouvert-local-llm-laptop-guide.md` | xpost | Paul Couvert: Best local models for laptop (Qwen3.6-27B, Gemma 4, Parakeet, Unsloth quants) |
--- a/wiki/log.md
+++ b/wiki/log.md
@ -2,6 +2,19 @@

 *Append-only changelog. Start: 2026-06-05*

+## [2026-06-30] Ingest | Paul Couvert — Best Local Models for Laptop (Cross-Platform Companion to MLX MoE)
+**Type:** ingest | **Scope:** raw/xpost, wiki/concepts/llm
+**Source:** X Post — https://x.com/itsPaulAi/status/2071717503105900993 (138 Likes, 12 Reposts, 18 Replies, 170 Bookmarks, ~13.5K Views)
+**Trigger:** Shared by Pit Weber in OME-Gruppe Topic "Tips & Tricks" (Topic 27) on 2026-06-29.
+**Actions:**
+- raw: `raw/xpost/2026-06-29_paulcouvert-local-llm-laptop-guide.md` (created — 4.9 KB; Frontmatter [type: xpost, author: @itsPaulAi, tags: local-llm, laptop, qwen, gemma, parakeet, unsloth, llama-cpp, lm-studio, quantization, voice-to-text, diffusion, edge-ai, google-ai-edge-gallery]. Content: Post summary, model overview table [5 models], practical tips [Unsloth, LM Studio, llama.cpp, Google AI Edge Gallery], engagement stats, Jun Song comparison table [5 dimensions], relevance to our setup [5 points], 6 cross-refs)
+- wiki (NEW): `concepts/llm/local-llm-laptop-guide.md` (created — 9.9 KB; Frontmatter [sources, tags]. Sektionen: Model Stack [5 model detail tables], Practical Deployment Tips [Unsloth, runtime options table], Resolving the Dense 27B Tension [context distinction table Paul vs Jun Song, resolution, recommendation], Complement to MLX MoE [comparison table], Model Landscape Summary [6-row table], Connection to Our Setup [model routing alignment, Cloud-Exit extension ladder table], 9 Cross-Refs, 5 external links)
+- wiki: `concepts/llm/mlx-moe-local-ai-optimization.md` (updated — cross-ref to new local-llm-laptop-guide.md added)
+- wiki: `index.md` (updated — Header auf "47. Update", neue LLM-Zeile für Local LLM Laptop Guide, neuer Raw-Sources-Eintrag)
+- log: this entry
+**Hector-Hauptthese:** Paul Couvert's Post ist der cross-platform Companion zu Jun Song's MLX MoE Post. Wo Jun Song die Apple Silicon / MLX-Perspektive liefert (MoE > Dense auf Mac), gibt Paul die "any laptop"-Perspektive (Unsloth quants + llama.cpp auf jeder Hardware). Der scheinbare Widerspruch (Paul empfiehlt Qwen3.6-27B dense, Jun Song warnt davor auf Mac) löst sich durch Kontext: Framework + Quantization machen den Unterschied. Die Konvergenz auf Qwen3.6-35B-A3B (MoE-Variante) ist bemerkenswert — beide empfehlen sie, Jun Song implizit (MoE-Präferenz), Paul explizit ("faster option"). Erweiterung der Cloud-Exit-Hardware-Ladder um die Laptop-Tier (kein Mac Studio/DGX/Strix Halo nötig). Parakeet 0.6B als lokaler STT-Kandidat ist relevant für OpenClaw Voice-Interaktion. Gemma 4 E4B auf Telefon erweitert Cloud-Exit auf Mobile und verbindet sich mit den neuen OpenClaw Mobile Apps.
+**Subagent-Modell:** ollama/glm-5.2:cloud
+
 ## [2026-06-30] Ingest | Jun Song — MLX MoE Optimization & The Local AI Info Gap
 **Type:** ingest | **Scope:** raw/xpost, wiki/concepts/llm
 **Source:** X Post — https://x.com/jun_song/status/2071625328083227118 (954 Likes, 60 Reposts, 51 Replies, 862 Bookmarks, ~89K Views)