hector-bot/knowledge-base
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

feat: DeRonin Chinese AI Stack — 87% cost-cut field report

Browse source 
- raw: raw/xpost/2026-06-29_deronin-chinese-ai-stack-cost-savings.md (NEW)
- wiki: concepts/llm/chinese-model-cost-routing.md (NEW — 8.5 KB)
  - Swap matrix: Opus→Kimi, GPT→Qwen, Sonnet→GLM, GPT-mini→MiMo, GPT-Image→Wan, Sora→Kling
  - Barbell routing connection + Hector-vs-DeRonin comparison
  - Factory-for-Gods thesis validation with empirical data
  - Cross-refs to 8 wiki pages + model-router skill
- wiki: index.md (updated — 43. Update)
- wiki: log.md (updated)
This commit is contained in:
Hector 2026-06-29 19:34:04 +02:00
parent fd39380f28
commit 22c5d016d7
4 changed files with 207 additions and 1 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

61
raw/xpost/2026-06-29_deronin-chinese-ai-stack-cost-savings.md Normal file
View file

@ -0,0 +1,61 @@
---
type: xpost
source_url: https://x.com/DeRonin_/status/2071561335234531578
author: DeRonin (@DeRonin_)
date: 2026-06-29
posted_by: Pit Weber (@PWeber)
topic: "Tips & Tricks (Topic 27)"
tags: [xpost, chinese-models, cost-routing, barbell, model-swap, kimi, qwen, glm, mimo, wan, kling, cost-reduction, ai-stack]
---
# DeRonin: "My entire AI stack is now Chinese 🇨🇳 — 87% cheaper, same revenue"
**Source:** [X-Post by @DeRonin_](https://x.com/DeRonin_/status/2071561335234531578)
**Posted:** 2026-06-29
**Shared by:** Pit Weber in OME-Gruppe, Topic "Tips & Tricks" (Topic 27)
## Post Content
> "My entire AI stack is now Chinese 🇨🇳 — 87% cheaper, same revenue"
DeRonin documents a full swap from Western frontier models to Chinese alternatives across six task categories, with a 30-day outcome report.
## Model Swaps by Task
| # | Task | Western Model → Chinese Model | Benchmark Gap | Cost Factor |
|---|------|-------------------------------|---------------|-------------|
| 1 | Reasoning / Backend Brain | Claude Opus 4.8 → **Kimi K2.7** | ~8% | ~11× cheaper |
| 2 | Code Generation | GPT-5.5 → **Qwen 3.7 Max** | ~18% | ~7× cheaper |
| 3 | Agent Loops + Tool Calling | Claude Sonnet 4.7 → **GLM 5.2** | ~3% | ~5× cheaper (input) |
| 4 | Cheap Volume / Bulk Processing | GPT-5.5 mini → **MiMo V2.5** | ~6% | ~12× cheaper |
| 5 | Image Generation | GPT-Image-2 → **Wan 2.5** | ~5% | ~8× cheaper |
| 6 | Video Generation | Sora 2 → **Kling 3.0** | roughly equal | ~6× cheaper |
## 30-Day Result
- **Operating costs:** dropped 87%
- **Output quality:** dropped 4% average
- **Revenue:** unchanged
- **Full article with routing logic:** promised for tomorrow (2026-06-30)
## Engagement
- 2.5K+ likes
- 130K+ views
## Relevance to Our Setup
This post directly validates our **Barbell Model Routing** decision (MEMORY.md, 2026-06-28):
- We already run **GLM 5.2** as primary (Task 3 in DeRonin's stack)
- **Kimi K2.7** is available as subagent option (Task 1)
- **Qwen3 Coder** and **MiMo** are in our fallback chain (Tasks 2 & 4)
- The `skills/openclaw-model-router` skill implements the 5-tier routing that DeRonin describes empirically
DeRonin's post is the **practical field report** for what Miles Deutscher formalized as "Token Engineering" (barbell strategy: expensive planning → cheap execution → expensive verification) and what TheProphet described macro-economically as "China builds the factory for gods" (see [[../../wiki/concepts/llm/ai-intelligence-commoditization-thesis.md]]).
## Related Wiki Pages
- `wiki/concepts/llm/chinese-model-cost-routing.md` — Wiki page created from this source
- `wiki/concepts/llm/ai-intelligence-commoditization-thesis.md` — Macro-thesis (Frontier vs. Factory)
- `wiki/architecture/model-routing.md` — Our routing architecture
- `skills/openclaw-model-router/SKILL.md` — 5-tier routing skill

132
wiki/concepts/llm/chinese-model-cost-routing.md Normal file
View file

@ -0,0 +1,132 @@
---
created: 2026-06-29
updated: 2026-06-29
sources:
- xpost/2026-06-29_deronin-chinese-ai-stack-cost-savings.md
- xpost/2026-06-28_milesdeutscher-token-engineering-barbell.md
tags: [concept, llm, chinese-models, cost-routing, barbell, model-swap, kimi, qwen, glm, mimo, wan, kling, cost-reduction, factory-for-gods, practical-report]
---
# Chinese Model Cost Routing — The 87% Cost-Cut Playbook
> **TL;DR:** DeRonin's 30-day field report swaps six Western frontier models for Chinese alternatives — 87% cost reduction, 4% quality drop, revenue unchanged. This is the empirical validation of the Barbell Model Routing strategy and TheProphet's "Factory for Gods" thesis.
## Sources
| Source | Author | Date | Angle |
|--------|--------|------|-------|
| [X-Post: "My entire AI stack is now Chinese"](https://x.com/DeRonin_/status/2071561335234531578) | DeRonin (@DeRonin_) | 2026-06-29 | Practical field report |
| Miles Deutscher "Token Engineering" | @Milesdeutscher | 2026-06-28 | Barbell strategy formalization |
| [TheProphet "Factory for Gods"](https://x.com/_The_Prophet__/status/2067390526157185188) | @TheProphet | 2026-06-17 | Macro-thesis: China industrializes intelligence |
## The Swap Matrix
DeRonin replaced every layer of his AI stack with Chinese models:
| Task | Western (before) | Chinese (after) | Quality Gap | Cost Factor | Notes |
|------|------------------|-----------------|-------------|-------------|-------|
| Reasoning / Backend Brain | Claude Opus 4.8 | **Kimi K2.7** (Moonshot) | ~8% | ~11× cheaper | Largest absolute saving — Opus is premium-priced |
| Code Generation | GPT-5.5 | **Qwen 3.7 Max** (Alibaba) | ~18% | ~7× cheaper | Biggest quality gap, but still "good enough" for production |
| Agent Loops + Tool Calling | Claude Sonnet 4.7 | **GLM 5.2** (Z.ai) | ~3% | ~5× cheaper (input) | Smallest gap — GLM 5.2 is near-parity for agentic tasks |
| Cheap Volume / Bulk | GPT-5.5 mini | **MiMo V2.5** (Xiaomi) | ~6% | ~12× cheaper | Best cost ratio in the stack |
| Image Generation | GPT-Image-2 | **Wan 2.5** | ~5% | ~8× cheaper | — |
| Video Generation | Sora 2 | **Kling 3.0** (Kuaishou) | ~equal | ~6× cheaper | Roughly equal quality at a fraction of cost |
### 30-Day Outcome
| Metric | Before | After | Delta |
|--------|--------|-------|-------|
| Operating costs | baseline | 13% of baseline | **−87%** |
| Output quality (avg) | baseline | 96% of baseline | **−4%** |
| Revenue | baseline | unchanged | **0%** |
**Key insight:** The 4% quality drop is across all tasks averaged. For agent loops (GLM 5.2) and video (Kling 3.0), the gap is ≤3% — effectively parity. The biggest gap (Qwen 3.7 Max for code gen at 18%) is offset by the fact that code generation volume is high and verification is cheap.
## Connection to Barbell Model Routing
DeRonin's stack is a **natural implementation of the Barbell Strategy** (Miles Deutscher, 2026-06-28):
| Barbell Phase | DeRonin's Implementation | Model |
|---------------|--------------------------|-------|
| **First 10%** (Planning / expensive) | Reasoning tasks where 8% quality gap matters | Kimi K2.7 (cheapest "brain") |
| **Middle 80%** (Execution / cheap) | Agent loops, bulk processing, code gen | GLM 5.2, MiMo V2.5, Qwen 3.7 Max |
| **Last 10%** (Verification / expensive) | Implicit — revenue unchanged means verification held | (not detailed in post) |
### Our Setup vs. DeRonin's
| Task | DeRonin | Hector (us) | Match? |
|------|---------|-------------|--------|
| Primary agent brain | Kimi K2.7 | GLM 5.2 (cloud) | Different — we route reasoning to GLM 5.2 |
| Code generation | Qwen 3.7 Max | Kimi K2.7 Code (subagent) | Different — we use Kimi for code |
| Agent loops + tools | GLM 5.2 | GLM 5.2 (native in OpenClaw v2026.6.8) | ✅ Same |
| Bulk processing | MiMo V2.5 | MiMo in fallback chain | ✅ Same family |
| Routing logic | (promised article) | `skills/openclaw-model-router` 5-tier | We have formalized routing |
**Where we differ:** DeRonin uses Kimi K2.7 for reasoning and GLM 5.2 for agent loops. We use GLM 5.2 as the primary for everything including reasoning, with Kimi K2.7 Code as a subagent option. His split is more granular — reasoning gets a dedicated "brain" model, agent loops get the cheaper agentic model. Our 5-tier router achieves similar granularity via task-complexity classification rather than task-type classification.
## Connection to "Factory for Gods" Thesis
This is the **micro-level proof** of TheProphet's macro-thesis (see [[ai-intelligence-commoditization-thesis.md]]):
> *"America has the frontier gods. China is building the factory for gods."*
DeRonin's 30-day report demonstrates empirically what TheProphet argued theoretically:
1. **Near-frontier is good enough** — A 4% average quality drop is invisible to end-users and revenue-neutral
2. **Cost asymmetry is extreme** — 5–12× cheaper per task category, compounding to 87% total
3. **The swap is not hypothetical** — It's a production stack that ran for 30 days with unchanged revenue
4. **Every layer is covered** — Not just text generation, but code, agents, images, and video
### The Quality-Cost Curve in Practice
TheProphet's thesis predicts that "a 10% intelligence gap is economically irrelevant when the second-best model is radically cheaper." DeRonin's data refines this:
- **Agent loops:** 3% gap at 5× cheaper → **no-brainer swap**
- **Reasoning:** 8% gap at 11× cheaper → **clearly worth it**
- **Code gen:** 18% gap at 7× cheaper → **the interesting edge case** — still profitable because code verification is cheaper than code generation
- **Video:** ~0% gap at 6× cheaper → **pure arbitrage**
The threshold is not "10% gap" universally — it's task-dependent. Where verification is cheap (code), larger gaps are tolerable. Where it's expensive (reasoning), smaller gaps matter more.
## Chinese Model Landscape (Referenced)
| Model | Vendor | Category | In Our Stack? |
|-------|--------|----------|---------------|
| Kimi K2.7 | Moonshot AI | Reasoning / coding | ✅ Subagent option |
| Qwen 3.7 Max | Alibaba | Code generation | Qwen3 Coder in fallback |
| GLM 5.2 | Z.ai | Agent loops / tool calling | ✅ Primary model |
| MiMo V2.5 | Xiaomi | Bulk / cheap volume | ✅ Fallback chain |
| Wan 2.5 | Alibaba (Wan) | Image generation | ❌ Not integrated |
| Kling 3.0 | Kuaishou | Video generation | ❌ Not integrated |
See [[llm-model-catalog.md]] for full model inventory and [[glm-5.2-zai-coding-model.md]] for our GLM 5.2 deep-dive.
## Implications for Our Architecture
1. **GLM 5.2 as primary is validated** — DeRonin independently chose GLM 5.2 for agent loops, the task category with the smallest quality gap (3%). Our default model choice is optimal for the highest-volume task type.
2. **Kimi K2.7 for reasoning is worth considering** — DeRonin splits reasoning (Kimi) from agent loops (GLM). Our current setup routes both to GLM 5.2. A potential optimization: route complex reasoning tasks (Tier 4 in our 5-tier router) to Kimi K2.7 instead of GLM 5.2.
3. **MiMo for bulk is confirmed** — We already have MiMo in the fallback chain. DeRonin's 12× cost factor for bulk processing validates Tier 0/1 routing.
4. **Image/Video gap to explore** — We don't currently use Wan 2.5 or Kling 3.0. If image/video generation becomes a regular need, these are the cost-optimal choices.
5. **Full article pending** — DeRonin promised a detailed article with routing logic for 2026-06-30. This page should be updated when it drops.
## Cross-References
- [[ai-intelligence-commoditization-thesis.md]] — TheProphet's "Factory for Gods" macro-thesis
- [[ai-value-migration-orchestration.md]] — Aravind's investor angle on value migration
- [[flat-curve-society.md]] — Yegge's "route to the dumbest model that can handle it"
- [[glm-5.2-zai-coding-model.md]] — GLM 5.2 deep-dive (our primary = DeRonin's agent loop choice)
- [[real-world-coding-showdown.md]] — Kimi K2.7 vs GLM 5.2 head-to-head
- [[llm-model-catalog.md]] — Full model inventory
- [[../../architecture/model-routing.md]] — Our routing architecture
- [[../policy/ai-as-geopolitical-weapon.md]] — Miles Deutscher's weaponization thesis
- `skills/openclaw-model-router/SKILL.md` — 5-tier routing skill (external to wiki)
## External Sources
- [DeRonin X-Post (Original)](https://x.com/DeRonin_/status/2071561335234531578)
- [Miles Deutscher "Token Engineering" (2026-06-28)](https://x.com/Milesdeutscher)
- [TheProphet "Factory for Gods" (2026-06-17)](https://x.com/_The_Prophet__/status/2067390526157185188)

3
wiki/index.md
View file

@ -2,7 +2,7 @@
*Auto-generated: 2026-06-23* *Auto-generated: 2026-06-23*
*Letzte Aktualisierung: 2026-06-29 (42. Update — Brian Roemmele Follow-Up Warning "I tried to warn ya": September-2024-Warnung als Prophetie bestätigt. Neue Raw-Datei `xpost/2026-06-28_roemmele-warning-open-source-ai-banned.md`. Aktualisiert: people/brian-roemmele.md, concepts/llm/reds-resilient-decentralized-swarm.md, concepts/policy/ai-regulation-2026.md.)* *Letzte Aktualisierung: 2026-06-29 (43. Update — DeRonin "My entire AI stack is now Chinese": 87% Kostenreduktion durch 6-Modell-Swap (Opus→Kimi, GPT→Qwen, Sonnet→GLM, GPT-mini→MiMo, GPT-Image→Wan, Sora→Kling). Validiert Barbell Routing + Factory-for-Gods-These. Neue Wiki-Seite `concepts/llm/chinese-model-cost-routing.md`.)*
## Architecture ## Architecture
@ -65,6 +65,7 @@
| [LLM Model Catalog](concepts/llm/llm-model-catalog.md) | Konsolidierte Modell-Übersicht aller im Wiki erwähnten LLMs. Fokus auf lokale Deployment-Optionen. Frontier-Tabelle (Cloud), Open-Source-Tabelle (HF-Links, Hosting, Praxis-Tests, Tester-Attribution), Hector's Active Stack, Post-Transformer-Outlook, DRACO-Fusion-Ergebnisse | 11 Wiki-Quellen + MEMORY.md | | [LLM Model Catalog](concepts/llm/llm-model-catalog.md) | Konsolidierte Modell-Übersicht aller im Wiki erwähnten LLMs. Fokus auf lokale Deployment-Optionen. Frontier-Tabelle (Cloud), Open-Source-Tabelle (HF-Links, Hosting, Praxis-Tests, Tester-Attribution), Hector's Active Stack, Post-Transformer-Outlook, DRACO-Fusion-Ergebnisse | 11 Wiki-Quellen + MEMORY.md |
| [LLM Sycophancy, Confabulation & Session-Statelessness — The Nano Banana Incident](concepts/llm/llm-sycophancy-confabulation.md) | Drei Mechanismen die LLM-Aussagen über eigene Fähigkeiten untrustworthy machen: Session-Statelessness, Sycophantic Compliance, Confabulation. Lüge-vs-Confabulation-vs-Sycophancy-Distinktion. Anti-Pattern: Modul-Aktivierung per Chat. Fallbeispiel: Gemini erfindet "Nano Banana 2". Goldene Regel: Provider-Doku > Chat | other/2026-06-26_nanobana-incident-sycophancy-confabulation.md | | [LLM Sycophancy, Confabulation & Session-Statelessness — The Nano Banana Incident](concepts/llm/llm-sycophancy-confabulation.md) | Drei Mechanismen die LLM-Aussagen über eigene Fähigkeiten untrustworthy machen: Session-Statelessness, Sycophantic Compliance, Confabulation. Lüge-vs-Confabulation-vs-Sycophancy-Distinktion. Anti-Pattern: Modul-Aktivierung per Chat. Fallbeispiel: Gemini erfindet "Nano Banana 2". Goldene Regel: Provider-Doku > Chat | other/2026-06-26_nanobana-incident-sycophancy-confabulation.md |
| [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 | | [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 |
### AGI ### AGI
| Seite | Beschreibung | Quellen | | Seite | Beschreibung | Quellen |

12
wiki/log.md
View file

@ -927,3 +927,15 @@ Bestehende `post-transformer-llm-architectures.md` bleibt als Vier-Säulen-Über
- log: this entry - log: this entry
**Hector-Hauptthese:** Das Berman-Video komplettiert das Hermes-Tutorial-Quartett Ende Juni 2026 (Jonas Keil Desktop-Review → Alex Finn v0.17.0 → Peter Yang Full Course → Berman 2-Min-Setup). Zusammen decken sie die volle Bandbreite ab: von Desktop-App über Feature-Deep-Dive über kompletten 45min-Kurs bis zum 2-Minuten-VPS-Quickstart. Die Hostinger-Sponsorship ist strategisch interessant: Hosting wird zum Klick-Produkt, Hermes positioniert sich als die "easy button"-Alternative zu OpenClaw's CLI-first-Ansatz. Die direkte OpenClaw-Vergleichung in einem Sponsored-Video unterstreicht die Konkurrenz-Narrative. **Hector-Hauptthese:** Das Berman-Video komplettiert das Hermes-Tutorial-Quartett Ende Juni 2026 (Jonas Keil Desktop-Review → Alex Finn v0.17.0 → Peter Yang Full Course → Berman 2-Min-Setup). Zusammen decken sie die volle Bandbreite ab: von Desktop-App über Feature-Deep-Dive über kompletten 45min-Kurs bis zum 2-Minuten-VPS-Quickstart. Die Hostinger-Sponsorship ist strategisch interessant: Hosting wird zum Klick-Produkt, Hermes positioniert sich als die "easy button"-Alternative zu OpenClaw's CLI-first-Ansatz. Die direkte OpenClaw-Vergleichung in einem Sponsored-Video unterstreicht die Konkurrenz-Narrative.
**Subagent-Modell:** ollama/glm-5.2:cloud **Subagent-Modell:** ollama/glm-5.2:cloud
## [2026-06-29] Ingest | DeRonin — "My entire AI stack is now Chinese 🇨🇳" (87% Cost-Cut Field Report)
**Type:** ingest | **Scope:** raw/xpost, wiki/concepts/llm, wiki/index
**Source:** X-Post by @DeRonin_ (https://x.com/DeRonin_/status/2071561335234531578) — shared by Pit Weber in OME-Gruppe Topic "Tips & Tricks" (Topic 27), 2026-06-29
**Trigger:** DeRonin dokumentiert einen vollständigen 6-Modell-Swap von Western frontier zu Chinese alternatives mit 30-Tage-Ergebnis: 87% Kostenreduktion, 4% Qualitätsverlust, Revenue unverändert. 2.5K+ likes, 130K+ views. Full Article mit Routing-Logic für 2026-06-30 angekündigt.
**Actions:**
- raw: `raw/xpost/2026-06-29_deronin-chinese-ai-stack-cost-savings.md` (created — 2.9 KB; Frontmatter mit type/source_url/author/date/posted_by/topic/tags, vollständige Swap-Matrix-Tabelle, 30-Day-Outcome, Engagement-Metriken, Relevance-to-Our-Setup mit Cross-Refs)
- wiki (NEU): `concepts/llm/chinese-model-cost-routing.md` (created — 8.5 KB; Umfassende Analyse-Seite mit 7 Sektionen: (1) Source-Übersicht mit 3 Quellen, (2) Swap-Matrix mit allen 6 Tasks + 30-Day-Outcome-Tabelle, (3) Barbell-Routing-Connection mit Phasen-Mapping + Hector-vs-DeRonin-Vergleichstabelle, (4) Factory-for-Gods-Connection mit TheProphet-These + empirischer Validierung, (5) Quality-Cost-Curve mit task-abhängigem Schwellwert, (6) Chinese Model Landscape mit In-Our-Stack-Marking, (7) Architecture Implications mit 5 konkreten Takeaways. Cross-Refs zu 8 Wiki-Seiten + model-router skill.)
- wiki: `index.md` (updated — Header auf "43. Update", neuer LLM-Eintrag unter Concepts/LLM Sektion)
- log: this entry
**Hector-Hauptthese:** DeRonin's Post ist das empirische Missing Link zwischen TheProphet's Makro-These ("Factory for Gods") und unserer Routing-Architektur. Während TheProphet die wirtschaftliche Logik erklärte und Miles Deutscher die Barbell-Strategie formalisierte, beweist DeRonin mit 30 Tagen Production-Daten: Der Swap ist nicht hypothetisch — er ist revenue-neutral. Die wichtigste Nuance: Der Qualitäts-Schwellwert ist task-abhängig. Bei Agent Loops (GLM 5.2, 3% Gap) ist es ein No-Brainer. Bei Code Gen (Qwen 3.7 Max, 18% Gap) funktioniert es nur, weil Code-Verifikation billiger ist als Code-Generierung. Das ist die "verification cost" als entscheidende Variable — ein Konzept das in unserer 5-Tier-Routing-Logik noch nicht explizit modelliert ist. DeRonin's unabhängige Wahl von GLM 5.2 für Agent Loops validiert unsere Default-Modell-Entscheidung. Sein Split (Kimi für Reasoning, GLM für Agent Loops) ist feingranularer als unser aktuelles Setup und könnte als Optimierungs-Input für den model-router skill dienen.
**Subagent-Modell:** ollama/glm-5.2:cloud