From 75f1ee69ac075bbbe309e3797fe02852c7dba92c Mon Sep 17 00:00:00 2001 From: JustVugg Date: Thu, 30 Jul 2026 09:05:00 +0200 Subject: [PATCH] docs: the README is about four model families, not one MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit The header still opened with "Explore GLM-5.2 (744B-parameter MoE)" while the repo description had already moved to "Run frontier MoE models". Inkling (975B) and Kimi K3 (2.8T) shipped in v1.3.0 and the roster table listing all four families sits at line 337 — so a reader met "GLM-5.2" at line 16 and only discovered the other three engines 300 lines later. Header now states the range rather than one model, and announces the roster up front with a link down to the table. A range does not go stale the next time someone ships a larger model; naming a single model does. The blockquote said "experimental inference engine and research platform" and then "not a production runtime with an SLA", which reads as a contradiction next to a usable engine that four model families run on. It is both things at once, and the reconciliation is a distinction rather than a hedge: no SLA on speed, a hard guarantee on semantics. That second half was already the last sentence of the paragraph — insufficient fast memory may reduce speed, it must not quietly redefine the model — it just was not where a reader would see it. Not touched, deliberately: every screenshot caption (744B at 4 tok/s, 19,456 experts, 13,260 characterised experts) describes a real run, and GLM-5.2 stays the worked example under "the core idea" because it is the model with the most measured data behind it. Reference model, not only model. Also fixes the sample banner, which claimed v1.1.0 across all four files. Co-Authored-By: Claude Opus 5 --- README.it.md | 39 +++++++++++++++++++++++---------------- README.md | 39 ++++++++++++++++++++++----------------- README.zh-CN.md | 22 +++++++++++++--------- README.zh-TW.md | 22 +++++++++++++--------- 4 files changed, 71 insertions(+), 51 deletions(-) diff --git a/README.it.md b/README.it.md index 945454679..f0d65fb55 100644 --- a/README.it.md +++ b/README.it.md @@ -7,25 +7,32 @@ English · 简体中文 · 繁體中文 · Italiano

-**Motore piccolo, modello immenso.** Esplora **GLM-5.2 (MoE da 744 miliardi di -parametri)** su hardware consumer ed eterogeneo — in C puro, senza dipendenze -del motore, trattando storage, RAM e VRAM come un'unica gerarchia di inferenza. - -> **Colibrì è un motore di inferenza sperimentale e una piattaforma di ricerca.** -> Il suo obiettivo principale è migliorare le prestazioni di inferenza lungo -> l'intero confine software/hardware — formati dei modelli, gerarchia di memoria, -> I/O dello storage, piazzamento, scheduling, kernel, speculazione e sovrapposizione -> CPU/GPU — affinché i grandi modelli dipendano meno da hardware raro e costino meno. - -Colibrì è intenzionalmente un luogo dove verificare idee di sistema aggressive, -non un runtime di produzione con SLA. Gli esperimenti devono dimostrare il proprio -valore con misure end-to-end riproducibili; la policy predefinita **non cambia mai -silenziosamente la precisione del modello né la semantica del router**. Una memoria -veloce insufficiente può ridurre la velocità, ma non ridefinire il modello di nascosto. +**Motore piccolo, modello immenso.** Esegui **modelli MoE di frontiera — da 744 +miliardi a 2,8 mila miliardi di parametri** — su hardware consumer ed eterogeneo, +in C puro e senza dipendenze del motore, trattando storage, RAM e VRAM come +un'unica gerarchia di inferenza. + +Oggi girano quattro famiglie: **GLM-5.2** (744B), **Inkling** (975B), **Kimi K3** +(2,8T) e **OLMoE** (7B) — un file C ciascuna, la stessa interfaccia `coli chat` / +`coli serve` / `coli web`. [Elenco completo](README.md#other-supported-models) + +> **Colibrì è un motore di inferenza che puoi usare oggi, e una piattaforma di +> ricerca aperta.** Il suo obiettivo principale è migliorare le prestazioni di +> inferenza lungo l'intero confine software/hardware — formati dei modelli, +> gerarchia di memoria, I/O dello storage, piazzamento, scheduling, kernel, +> speculazione e sovrapposizione CPU/GPU — affinché i grandi modelli dipendano +> meno da hardware raro e costino meno. + +Colibrì è intenzionalmente un luogo dove verificare idee di sistema aggressive — +quindi **nessuno SLA sulla velocità, e una garanzia dura sulla semantica**: gli +esperimenti devono dimostrare il proprio valore con misure end-to-end riproducibili; +la policy predefinita **non cambia mai silenziosamente la precisione del modello né +la semantica del router**. Una memoria veloce insufficiente può ridurre la velocità, +ma non ridefinire il modello di nascosto. ``` $ ./coli chat - 🐦 colibri v1.1.0 — GLM-5.2 · 744B MoE · int4 · streaming CPU + 🐦 colibri v1.3.0 — GLM-5.2 · 744B MoE · int4 · streaming CPU ✓ ready in 32s · resident 9.9 GB › ciao! ◆ Ciao! 😊 Come posso aiutarti oggi? diff --git a/README.md b/README.md index ff4a7bcab..7796b3f9a 100644 --- a/README.md +++ b/README.md @@ -13,26 +13,31 @@ English · 简体中文 · 繁體中文 · Italiano

-**Tiny engine, immense model.** Explore **GLM-5.2 (744B-parameter MoE)** across -consumer and heterogeneous hardware — in pure C, with zero engine dependencies, -by treating storage, RAM, and VRAM as one inference hierarchy. - -> **Colibrì is an experimental inference engine and research platform.** Its -> primary goal is to pursue inference-side performance across the entire -> software/hardware boundary — model formats, memory hierarchy, storage I/O, -> placement, scheduling, kernels, speculation, and CPU/GPU overlap — so large -> models depend less on scarce hardware and cost less to run. - -Colibrì treats VRAM, RAM, and storage as one managed memory hierarchy. It is -deliberately a place to test aggressive systems ideas, not a production runtime -with an SLA. Experiments must earn their place through reproducible end-to-end -measurements, and the default policy **never silently changes model precision -or router semantics**. Insufficient fast memory may reduce speed; it must not -quietly redefine the model. +**Tiny engine, immense model.** Run **frontier MoE models — 744B to 2.8T +parameters** — on consumer and heterogeneous hardware, in pure C with zero +engine dependencies, by treating storage, RAM, and VRAM as one inference +hierarchy. + +Four families run today: **GLM-5.2** (744B), **Inkling** (975B), **Kimi K3** +(2.8T) and **OLMoE** (7B) — one C file each, the same `coli chat` / +`coli serve` / `coli web` front end. [Full roster ↓](#other-supported-models) + +> **Colibrì is an inference engine you can run today, and an open research +> platform.** Its primary goal is to pursue inference-side performance across +> the entire software/hardware boundary — model formats, memory hierarchy, +> storage I/O, placement, scheduling, kernels, speculation, and CPU/GPU +> overlap — so large models depend less on scarce hardware and cost less to run. + +Colibrì treats VRAM, RAM, and storage as one managed memory hierarchy, and it is +deliberately a place to test aggressive systems ideas — so there is **no SLA on +speed, and a hard guarantee on semantics**: experiments must earn their place +through reproducible end-to-end measurements, and the default policy **never +silently changes model precision or router semantics**. Insufficient fast memory +may reduce speed; it must not quietly redefine the model. ``` $ ./coli chat - 🐦 colibri v1.1.0 — GLM-5.2 · 744B MoE · int4 · streaming CPU + 🐦 colibri v1.3.0 — GLM-5.2 · 744B MoE · int4 · streaming CPU ✓ ready in 32s · resident 9.9 GB › ciao! ◆ Ciao! 😊 Come posso aiutarti oggi? diff --git a/README.zh-CN.md b/README.zh-CN.md index 47d3136fc..84c87ae71 100644 --- a/README.zh-CN.md +++ b/README.zh-CN.md @@ -7,20 +7,24 @@ English · 简体中文 · 繁體中文 · Italiano

-**小巧引擎，庞大模型。**在消费级与异构硬件上探索 **GLM-5.2（744B 参数的 -MoE）**——以引擎零依赖的纯 C 实现，将存储、RAM 与 VRAM 视为统一的推理层级。 +**小巧引擎，庞大模型。**在消费级与异构硬件上运行**前沿 MoE 模型——从 744B 到 +2.8T 参数**——以引擎零依赖的纯 C 实现，将存储、RAM 与 VRAM 视为统一的推理层级。 -> **Colibrì 是一个实验性推理引擎与研究平台。**它的首要目标是在完整的软硬件边界上 -> 追求推理侧性能——模型格式、内存层级、存储 I/O、放置、调度、内核、推测解码以及 -> CPU/GPU 重叠执行——让大模型减少对稀缺硬件的依赖，并降低运行成本。 +目前可运行四个模型家族：**GLM-5.2**（744B）、**Inkling**（975B）、**Kimi K3** +（2.8T）与 **OLMoE**（7B）——各自一个 C 文件，共用同一套 `coli chat` / +`coli serve` / `coli web` 前端。[完整列表](README.md#other-supported-models) -Colibrì 刻意用于验证激进的系统思路，而不是提供 SLA 的生产运行时。实验必须通过 -可复现的端到端测量证明价值；默认策略**绝不会在未告知的情况下改变模型精度或路由语义**。 -高速内存不足可以降低速度，但不能悄悄重新定义模型。 +> **Colibrì 既是今天就能运行的推理引擎，也是一个开放的研究平台。**它的首要目标是在 +> 完整的软硬件边界上追求推理侧性能——模型格式、内存层级、存储 I/O、放置、调度、内核、 +> 推测解码以及 CPU/GPU 重叠执行——让大模型减少对稀缺硬件的依赖，并降低运行成本。 + +Colibrì 刻意用于验证激进的系统思路——因此**对速度不作 SLA 承诺，对语义则给出硬性保证**： +实验必须通过可复现的端到端测量证明价值；默认策略**绝不会在未告知的情况下改变模型精度或 +路由语义**。高速内存不足可以降低速度，但不能悄悄重新定义模型。 ``` $ ./coli chat - 🐦 colibri v1.1.0 — GLM-5.2 · 744B MoE · int4 · streaming CPU + 🐦 colibri v1.3.0 — GLM-5.2 · 744B MoE · int4 · streaming CPU ✓ ready in 32s · resident 9.9 GB › ciao! ◆ Ciao! 😊 Come posso aiutarti oggi? diff --git a/README.zh-TW.md b/README.zh-TW.md index 815f931da..bb8bdbc6d 100644 --- a/README.zh-TW.md +++ b/README.zh-TW.md @@ -7,20 +7,24 @@ English · 简体中文 · 繁體中文 · Italiano

-**小巧引擎，龐大模型。**在消費級與異質硬體上探索 **GLM-5.2（744B 參數的 -MoE）**——以引擎零相依套件的純 C 實作，將儲存、RAM 與 VRAM 視為統一的推論階層。 +**小巧引擎，龐大模型。**在消費級與異質硬體上執行**前沿 MoE 模型——從 744B 到 +2.8T 參數**——以引擎零相依套件的純 C 實作，將儲存、RAM 與 VRAM 視為統一的推論階層。 -> **Colibrì 是一個實驗性推論引擎與研究平台。**它的首要目標是在完整的軟硬體邊界上 -> 追求推論側效能——模型格式、記憶體階層、儲存 I/O、配置、排程、核心、推測解碼以及 -> CPU/GPU 重疊執行——讓大型模型減少對稀缺硬體的依賴，並降低執行成本。 +目前可執行四個模型家族：**GLM-5.2**（744B）、**Inkling**（975B）、**Kimi K3** +（2.8T）與 **OLMoE**（7B）——各自一個 C 檔案，共用同一套 `coli chat` / +`coli serve` / `coli web` 前端。[完整清單](README.md#other-supported-models) -Colibrì 刻意用於驗證激進的系統構想，而不是提供 SLA 的生產執行環境。實驗必須透過 -可重現的端到端測量證明價值；預設策略**絕不會在未告知的情況下改變模型精度或路由語意**。 -高速記憶體不足可以降低速度，但不能悄悄重新定義模型。 +> **Colibrì 既是今天就能執行的推論引擎，也是一個開放的研究平台。**它的首要目標是在 +> 完整的軟硬體邊界上追求推論側效能——模型格式、記憶體階層、儲存 I/O、配置、排程、核心、 +> 推測解碼以及 CPU/GPU 重疊執行——讓大型模型減少對稀缺硬體的依賴，並降低執行成本。 + +Colibrì 刻意用於驗證激進的系統構想——因此**對速度不作 SLA 承諾，對語意則給出硬性保證**： +實驗必須透過可重現的端到端測量證明價值；預設策略**絕不會在未告知的情況下改變模型精度或 +路由語意**。高速記憶體不足可以降低速度，但不能悄悄重新定義模型。 ``` $ ./coli chat - 🐦 colibri v1.1.0 — GLM-5.2 · 744B MoE · int4 · streaming CPU + 🐦 colibri v1.3.0 — GLM-5.2 · 744B MoE · int4 · streaming CPU ✓ ready in 32s · resident 9.9 GB › ciao! ◆ Ciao! 😊 Come posso aiutarti oggi?