GitHub - quantumaikr/quant.cpp: LLM inference with 7x longer context. Pure C, zero dependencies. Lossless KV cache compression + single-header library.

The single-header C engine for KV-compressed LLM inference

Production: uniform_4b KV cache (4–7x compression at +6% PPL on Llama 3.2 3B).
Research: building blocks for TurboQuant, PolarQuant, QJL — 7 KV quantization types in one engine.
72K LOC pure C, zero dependencies. Ships as quant.h — drop one file into any project.
Runs everywhere a C compiler does: iOS · Android · WASM · MSVC · microcontrollers.

The Problem

LLM memory is dominated by the KV cache, not model weights. At 32K context, a 8B model's KV cache consumes 4GB — more than the model itself. Every existing engine stores KV in FP16. We compress it.

  +------------+-------------------------------+
  |            | KV Cache (FP16)               |
  | Model(4GB) | ██████████████   8K  <-- OOM  |
  +------------+-------------------------------+
  |            | KV (4-bit)                    |
  | Model(4GB) | ██ -------------> 350K ctx    |
  |            |      6.9x smaller             |
  +------------+-------------------------------+

The Result

Same hardware. 4–7x longer context. PPL measured and disclosed.

Llama 3.2 3B Instruct — PPL × Speed (FP32 KV NEON = 13.56 PPL @ 14.8 tok/s)

9 rounds of Karpathy iteration closed the quant-KV speed gap to FP32 KV from −45% to −8%, while delivering 5.8–7.1× memory compression. We do not (yet) beat fp32 in raw speed, but we get within 8% of it for ~7× less memory.

KV Config	Bytes/block	Compression	PPL	Δ vs FP32	tok/s	vs FP32 speed
FP32 reference (NEON)	—	1×	13.56	—	14.83	baseline
`turbo_kv_5b` 🏆 quality	88	5.8×	13.65	+0.7%	13.13	−11.5%
`turbo_kv_4bo` 🧪	96	5.3×	13.90	+2.5%	12.7	−14%
`turbo_kv_4b` ⭐ default	72	7.1×	14.33	+5.7%	13.67	−7.8%
`turbo_kv_3b`	56	9.1×	15.36	+13.3%	13.4	−9.6%
`turbo_kv_3bo` 🧪	80	6.4×	14.17	+4.5%	9.3	−37%
`uniform_4b`	68	7.5×	14.60	+7.7%	11.7	−21%
llama.cpp `q4_0` KV (lit.)	~70	~7.3×	~14.99	+10.6%	—	—

                  PPL Degradation vs FP32           Speed vs FP32 KV
                       (lower is better)              (higher is better)

  turbo_kv_5b     │█ +0.7%                       █████████ −11.5%
  turbo_kv_4bo    │██▌ +2.5%                     ████████ −14%
  turbo_kv_4b ⭐  │█████ +5.7%                   ██████████ −7.8%
  turbo_kv_3b     │█████████████ +13.3%          █████████ −9.6%
  uniform_4b      │██████ +7.7%                  ███████ −21%
  llama.cpp q4_0  │██████████ +10.6%             — (not measured)
  FP32 reference  │ ← 0%                         14.83 tok/s ←
                   0%   +5%   +10%               0   25%   50%   75%   100%

turbo_kv_4b (default) and turbo_kv_5b (quality) are the Pareto-optimal recommendations: 5.8–7.1× memory compression at 92% of FP32 KV speed. Full Karpathy-loop history (9 rounds across 3 sessions) in bench/results/turboquant_reproduction.md.

Cross-size validation (3 Llama-family models)

Model	turbo_kv_5b PPL Δ	turbo_kv_4b PPL Δ
SmolLM2 135M Instruct	+1.7%	+5.8%
Llama 3.2 1B Instruct	+0.7%	+7.3%
Llama 3.2 3B Instruct	+0.7%	+5.7%

turbo_kv_5b is consistently near-lossless across model sizes (~1% PPL Δ). turbo_kv_4b stays in the 5–8% range. Recommendation: use turbo_kv_3b only on models ≥ 3B parameters (the 8-level codebook is too coarse for small models — +61% PPL on Llama 3.2 1B).

About this comparison: We previously published v0.6.3 release notes claiming turbo_kv beats fp32 KV speed. That was an artifact of the fp32 attention path being unoptimized scalar — once we added NEON to the fp32 path (commit 4490c83), the honest gap is −7% to −12%, not +5% to +10%. We've corrected the README and the v0.6.3 release notes.

Context length gains (`turbo_kv_4b` + `q4` value cache)

Hardware	Model	FP16 KV ctx	quant.cpp ctx	KV Gain
16GB Mac	Llama 3.2 3B	50K tokens	350K tokens	6.9x
16GB Mac	Gemma 4 26B MoE	4K tokens	14K tokens	3.5x
8GB Laptop	Llama 8B (Q4)	16K tokens	61K tokens	3.8x
24GB RTX 3090	Llama 8B (Q4)	147K tokens	559K tokens	3.8x

Why quant.cpp?

LLM memory is dominated by the KV cache. quant.cpp is a minimal C engine that ships KV cache quantization that actually works, in a form factor nobody else offers: one single header, zero dependencies, runs on iOS/Android/WASM/MSVC/microcontrollers.

Two reasons to use it:

You need to embed LLM inference inside something. An app, a game, a web page, a device. quant.cpp is one file (quant.h, 628KB) plus libc. Everywhere a C compiler runs, this runs.
You want to study KV cache compression. quant.cpp implements 7 KV quantization schemes side by side: uniform_4b/2b/3b, polar_3b/4b, qjl_1b, turbo_kv_*. You can read each one in a single C file and add a new one in 3 functions.

Honest disclosure: In April 2026 Google published TurboQuant (ICLR 2026). quant.cpp's turbo_kv_* types started as a port of that algorithmic structure (Random Hadamard Transform → Lloyd-Max codebook → 1-bit QJL residual). Through a Karpathy-loop ablation we discovered the QJL residual stage was contributing literally zero to scores, dropped it, and reinvested the freed bytes into a larger codebook. The result (turbo_kv_4b at 14.28 PPL on Llama 3.2 3B) beats our previous production champion uniform_4b and llama.cpp's q4_0 KV at the same 4-bit budget. The full optimization history is in bench/results/turboquant_reproduction.md.

Need the exact paper numbers in a paper? Use Google's reference. Need a small, readable C engine with KV compression that ships on a phone, browser, microcontroller, or game engine? Use quant.cpp.

Get Started in 60 Seconds

# 1. Build
git clone https://github.com/quantumaikr/quant.cpp && cd quant.cpp
cmake -B build -DCMAKE_BUILD_TYPE=Release && cmake --build build -j$(nproc)

# 2. Download a model (135MB starter)
pip install huggingface_hub
hf download bartowski/SmolLM2-135M-Instruct-GGUF SmolLM2-135M-Instruct-Q8_0.gguf --local-dir models/

# 3. Run
./build/quant models/SmolLM2-135M-Instruct-Q8_0.gguf --chat -p "Hello!" -j 4

# 4. With KV compression (7x longer context)
./build/quant models/SmolLM2-135M-Instruct-Q8_0.gguf --chat -p "Hello!" -k uniform_4b -v q4

Full API docs · WASM demo · Add your own KV type · Python: pip install quantcpp

See It In Action: Book-in-a-Chat

Load an entire novel into context and ask questions about it. llama.cpp runs out of memory. quant.cpp remembers the whole book.

# Load Alice in Wonderland (~27K tokens) with KV compression
bash bench/demo/book_chat.sh models/Llama-3.2-3B-Instruct-Q8_0.gguf

# Q: "What riddle did the Mad Hatter ask Alice?"
# A: "Why is a raven like a writing-desk?" — from Chapter 7, A Mad Tea-Party...

On a 16GB Mac with Llama 3.2 3B: llama.cpp maxes out at ~50K tokens (FP16 KV). quant.cpp compresses KV 6.9x → 350K tokens — enough for 12 novels.

How It Compares

vs llama.cpp: Quality at same bit budget

                    KV Quantization Quality (SmolLM2 1.7B, WikiText-2)
                    
  llama.cpp Q4_0 KV │██████████████████████████████████████ PPL +10.6%
                    │
  llama.cpp Q8 K+Q5 V │▎ PPL ~+1%  ← recommended (1.6x compression)
                    │
   quant.cpp 4-bit  │▏ PPL +0.0%  ← lossless (3.8x compression)
                    │
   quant.cpp 3-bit  │█ PPL +1.3%  ← delta compression (4.3x)
                    └────────────────────────────────────────────────
                     0%                                         +12%
                              Perplexity Degradation →

Both are per-block methods. The quality gap comes from block size (128 vs 32), min-max range encoding, independent K/V treatment, and delta compression — not from a fundamental design flaw in llama.cpp. At ~1.6x compression, llama.cpp Q8+Q5 is excellent. quant.cpp targets the 4-7x range where the difference matters.

vs other TurboQuant implementations

	quant.cpp	turbo-quant (Rust)	turboquant-pytorch	scos-lab/turboquant
Language	Pure C11	Rust	Python	Python
Single-header	✅ quant.h (628KB)	❌ Cargo crate	❌ pip install	❌
Dependencies	libc + libm	Rust toolchain	PyTorch + CUDA	PyTorch
iOS / Android	✅	❌	❌	❌
WASM (browser)	✅ 192KB	❌	❌	❌
MCU / embedded	✅	❌	❌	❌
Windows MSVC	✅	✅	(Python)	(Python)
GGUF model loading	✅ 7 architectures	❌	❌	research only
End-to-end inference	✅	kernel only	kernel only	kernel only

vs production inference engines

	quant.cpp	llama.cpp	vLLM	MLX
KV quantization	TurboQuant + 6 schemes	Q8_0/Q5_0 (2x)	--	--
Code size	72K LOC	250K+	100K+	50K+
Embeddable	single header	library	library	framework
Read in an afternoon	✅	❌	❌	❌
GPU throughput	basic	full	best	Metal

Use llama.cpp for speed on a workstation. Use vLLM for batch serving. Use quant.cpp when you need to ship LLM inference inside something — an app, a game, a website, a device.

Supported Models

Model	Params	Architecture	Speed (M1 Pro, 8T)	KV Compression
SmolLM2 135M	135M	Llama	103 tok/s	2.4x
Llama 3.2 3B Instruct	3B	Llama 3 (GQA)	10 tok/s	6.9x
Gemma 4 26B-A4B-it	26B (4B active)	MoE 128 experts	3.9 tok/s	3.5x
Qwen3.5 0.8B	752M	DeltaNet hybrid	80 tok/s	3.8x
Qwen3.5 4B	4B	DeltaNet hybrid	20 tok/s	3.8x
SmolLM2 1.7B	1.7B	Llama	25 tok/s	3.8x
Gemma 3 270M	270M	Gemma 3	176 tok/s	3.8x

GGUF format. Load any llama.cpp-compatible model.

Gemma 4 26B-A4B architecture details

Full support for Gemma 4's hybrid MoE architecture:

Dual-FFN: parallel Dense MLP + 128-expert MoE per layer
Hybrid attention: 25 sliding (head_dim=256) + 5 full (head_dim=512) layers
QK-norm aware KV compression: auto FP32 keys + Q4 values (3.5x savings)
Learned RoPE with per-layer frequency factors
IQ3_XXS/IQ4_NL fused dot with NEON optimization for MoE experts
GeGLU activation (NEON-accelerated fast tanh approximation)

./build/quant gemma-4-26B-A4B-it-UD-Q3_K_M.gguf \
  -p "<start_of_turn>user\nWhat is the capital of France?\n<end_of_turn>\n<start_of_turn>model\n" \
  -n 50 -j 8 -T 0.0 -k uniform_4b -v q4
# Output: "The capital of France is **Paris**."

KV Cache Compression

The Idea

Standard:  Store every key as-is            → 16 bits/element → FP16

quant.cpp: Quantize keys to 4-bit           → 4 bits/element  → 3.8x
           + quantize values to Q4           → 4 bits/element  → 6.9x
           + delta encode adjacent keys      → 3 bits/element  → 8.5x

Like video compression: I-frames (FP32) every 64 tokens, P-frames (3-bit delta) between.

Quality vs Compression

                    WikiText-2 PPL (SmolLM2 1.7B)

  FP32 baseline      14.63 │ ●
  4b K + FP16 V       14.63 │ ● identical
  4b K + Q4 V         14.57 │ ● slightly better (!)
  delta 3b K + Q4 V   14.82 │  ●  +1.3%
  llama.cpp Q8K+Q5V   ~14.8 │  ●  ~+1% (1.6x compression)
  llama.cpp Q4_0 KV   16.18 │          ● +10.6% (3.8x compression)
  3b K (no delta)       ——  │                              ● +62%
                            └──┬──┬──┬──┬──┬──┬──┬──┬──┬──┬──
                              14  15  16  17  18  19  20  21+

Modes

Config	Compression	PPL vs FP32	Best for
`delta + 3b K + Q4 V`	~8.5x	+1.3%	Maximum context
`delta + 4b K + Q4 V`	~6.9x	~0%	Quality + compression
`uniform_4b K + Q4 V`	6.9x	~0%	Simple, no delta overhead
`uniform_4b K + FP16 V`	1.6x	+0.0%	Lossless baseline

QK-norm Aware (Gemma 4)

Models with QK-norm normalize keys to the unit sphere, creating extremely sparse distributions. quant.cpp auto-detects this and stores keys in FP32 while quantizing only values — preserving perfect precision with 3.5x V memory reduction.

Advanced Usage

# Delta compression (maximum context, 8.5x)
./build/quant model.gguf --chat -p "hello" -k uniform_3b -v q4 --delta

# Perplexity benchmark
./build/quant model.gguf --ppl input.txt -k uniform_4b -v q4

# Model info
./build/quant model.gguf --info

# Performance profiling
./build/quant model.gguf --chat -p "hello" -n 50 --profile

Single-Header Mode

Copy one file. Add LLM to any C project.

#define QUANT_IMPLEMENTATION
#include "quant.h"

int main() {
    quant_model* m = quant_load("model.gguf");
    quant_ctx*   c = quant_new(m, NULL);
    
    // Streaming
    quant_generate(c, "Tell me a joke", print_token, NULL);
    
    // Or one-shot
    char* answer = quant_ask(c, "What is 2+2?");
    printf("%s\n", answer);
    free(answer);
    
    quant_free_ctx(c);
    quant_free_model(m);
}

cc app.c -o app -lm -lpthread    # that's it — no cmake, no framework

15.7K LOC, 643KB, ~2s compile time. Full API:

Function	Description
`quant_load(path)`	Load a GGUF model
`quant_new(model, config)`	Create inference context
`quant_generate(ctx, prompt, cb, ud)`	Stream tokens via callback
`quant_ask(ctx, prompt)`	Generate and return string
`quant_free_ctx(ctx)`	Free context
`quant_free_model(model)`	Free model

Browser Demo (WASM)

192KB. The entire inference engine compiles to a WASM binary smaller than most JPEGs.

cd wasm && bash build.sh          # Requires: emscripten
python3 -m http.server 8080       # Serve locally
# Open http://localhost:8080, drag & drop any GGUF model

Everything runs client-side. Nothing is uploaded. KV compression active by default.

Docker & Server

Docker (zero-dependency, ~10MB image):

docker build -t quant.cpp .
docker run -v ./models:/models quant.cpp /models/model.gguf -p "hello" -k uniform_4b -v q4

OpenAI-compatible server (/v1/chat/completions):

cmake -B build -DTQ_BUILD_SERVER=ON && cmake --build build
./build/quant-server model.gguf -p 8080 -k uniform_4b

# Works with the OpenAI Python SDK
curl http://localhost:8080/v1/chat/completions \
  -d '{"messages":[{"role":"user","content":"Hello"}],"max_tokens":64}'

Build with -DTQ_BUILD_SERVER=ON. Streaming SSE supported. KV compression configurable per request.

Python

cd bindings/python && pip install .

from quantcpp import Model

with Model("model.gguf", kv_compress=1) as m:
    print(m.ask("What is the capital of France?"))

    # Streaming
    for token in m.generate("Once upon a time"):
        print(token, end="", flush=True)

Zero build dependencies beyond a C compiler. Compiles quant.h at install time.

Backends & Performance

Backend	Platform	Status	Notes
NEON	ARM (Apple Silicon)	Production	5.8x SIMD speedup
AVX2	x86	Production
Metal	Apple GPU	Verified	Batch matmul dispatch
CUDA	NVIDIA GPU	Compiles
Vulkan	Cross-platform	Compiles
WASM	Browser	NEW	192KB binary
MSVC	Windows	NEW	VS 2019/2022

Performance breakdown (Gemma 4 26B on M1 Pro)

Component	ms/token	Share
Attention matmul (Q8_0 NEON)	168	65%
MoE experts (IQ3_XXS/IQ4_NL NEON)	72	28%
Attention scores	3	1%
Other	14	6%
Total	257	3.9 tok/s

FAQ

How is this different from llama.cpp?

llama.cpp is a full-featured inference framework (250K+ LOC). quant.cpp is a minimal engine (72K LOC) you can read, modify, and embed. Different tools for different problems: llama.cpp optimizes speed, quant.cpp optimizes memory (KV compression) and embeddability (single header).

llama.cpp already has KV quantization. How is yours different?

llama.cpp supports KV cache quantization (Q8_0 K + Q5_0 V is the recommended config, ~1.6x compression with minimal quality loss). quant.cpp targets higher compression: 4-bit K + Q4 V gives 3.8x at +0.0% PPL, and delta compression pushes to 4.3x at +1.3% PPL. The quality advantage comes from 128-element min-max blocks (vs 32-element), independent K/V quantization methods, and delta encoding of adjacent keys — a technique llama.cpp doesn't have. Use llama.cpp's KV quant if 1.6x is enough; use quant.cpp if you need 4-7x.

How does this compare to Karpathy's llm.c?

Similar philosophy: minimal C, educational. Key differences: quant.cpp supports quantized weights (Q4_K_M, Q8_0, IQ2), multiple architectures (Llama, Qwen, Gemma, MoE), GGUF loading, and KV cache compression. Think of llm.c as the textbook and quant.cpp as the production-ready version.

Can I embed this in my app?

Yes. Two options:

Single-header: Copy quant.h, #define QUANT_IMPLEMENTATION in one .c file. Done.
Full library: Link against libturboquant.a.

Works on Linux, macOS, Windows (MSVC/MinGW), iOS, Android, and WASM.

Why is it slower than llama.cpp?

Three reasons: (1) llama.cpp has years of hand-tuned NEON/AVX2 assembly for every quant format, (2) llama.cpp offloads the full forward pass to Metal/CUDA GPU, (3) 250K+ LOC vs 72K LOC means more micro-optimizations. quant.cpp optimized for memory and embeddability first. Speed improvements (full Metal GPU offload, more SIMD kernels) are actively in progress — see v1.3 plan.

No GPU — is this useless?

If you need 100+ tok/s, use llama.cpp with Metal/CUDA. If you need to embed inference in an iOS app, WASM module, game engine, or IoT device — quant.cpp works. CPU on Apple Silicon: 25 tok/s (1.7B), 11.6 tok/s (3B), 3.9 tok/s (26B MoE).

Can it run in the browser?

Yes. cd wasm && bash build.sh. The WASM binary is 192KB. Drop a GGUF model and chat. Everything runs client-side.

What about sub-3-bit quantization?

Tested extensively (2-bit delta, NF2, online SVD, multi-hash). None reached acceptable quality. Per-step cosine 0.997 compounds to 0.885 after 200 steps. 3-bit + delta is the practical minimum.

Documentation

Document	Description
API Reference	Full C API for quant.h and libturboquant (730 lines)
Custom Quantization	Add your own KV type in 3 functions
H2H Benchmark	Reproducible quant.cpp vs llama.cpp comparison
KV Compression Landscape	Eviction vs Architecture vs Compression guide
ROADMAP	Project direction and planned features
CHANGELOG	Version history and release notes
Tech Report	Architecture and benchmarks (Arxiv draft)
WASM Demo	Try it in your browser — no install needed

References & Citations

quant.cpp is an independent implementation of published research. The Variant F architecture (RHT preprocessing + scalar Lloyd-Max codebook on rotated values, no QJL stage) sits in a lineage that combines two prior works:

HIGGS — Malinovskii, Panferov, Ilin, Guo, Richtárik, Alistarh. Pushing the Limits of Large Language Model Quantization via the Linearity Theorem. Nov 2024. arXiv:2411.17525. HIGGS introduced the Random Hadamard Transform + MSE-optimal grid quantization pattern (for weight quantization). Our tq_rht.c Walsh-Hadamard + Rademacher implementation follows this pattern. Credit to Tim Dettmers (discussion thread) for pointing this out.
TurboQuant — Zandieh, Daliri, Hadian, Mirrokni. TurboQuant: Online Vector Quantization with Near-optimal Distortion Rate. ICLR 2026. arXiv:2504.19874. TurboQuant applies the rotation pattern to KV cache with a 1-bit QJL residual stage and per-channel outlier handling. Our work started as a literal port of TurboQuant; through 9 rounds of Karpathy-loop iteration we simplified it (dropped QJL, dropped outlier channels) into the current Variant F. We do not claim our shipped variant is the TurboQuant algorithm — it is an empirically-derived simplification.
PolarQuant — Quantizing KV Caches with Polar Transformation. AISTATS 2026. arXiv:2502.02617. The polar-coordinate KV quantization that our tq_polar.c baseline implements.
QJL — Quantized Johnson-Lindenstrauss Transform for KV Cache Compression. AAAI 2025. arXiv:2406.03482. The 1-bit sketch building block. Used in our tq_qjl.c baseline; we found it contributed ~zero to attention scores in the Variant F regime and dropped it.
Google Research blog post on TurboQuant

Honest attribution: Variant F's structure (RHT + scalar grid quantization) is closest to HIGGS in spirit, applied to KV cache like TurboQuant, with both the QJL residual and the outlier channel split removed through ablation. If you use quant.cpp in academic work, please cite all three (HIGGS, TurboQuant, PolarQuant) and this repository.

QuantumAI · GitHub

Name		Name	Last commit message	Last commit date
Latest commit History 302 Commits
.claude		.claude
.github		.github
bench		bench
bindings/python		bindings/python
docs		docs
examples		examples
harness		harness
include/turboquant		include/turboquant
integrations		integrations
scripts		scripts
spec		spec
src		src
tests		tests
tools		tools
tq_convert.dSYM/Contents		tq_convert.dSYM/Contents
tq_run.dSYM/Contents		tq_run.dSYM/Contents
wasm		wasm
.dockerignore		.dockerignore
.gitignore		.gitignore
CHANGELOG.md		CHANGELOG.md
CLAUDE.md		CLAUDE.md
CMakeLists.txt		CMakeLists.txt
CODE_OF_CONDUCT.md		CODE_OF_CONDUCT.md
CONTRIBUTING.md		CONTRIBUTING.md
Dockerfile		Dockerfile
LICENSE		LICENSE
Makefile		Makefile
README.ko.md		README.ko.md
README.md		README.md
ROADMAP.md		ROADMAP.md
SECURITY.md		SECURITY.md
docker-compose.yml		docker-compose.yml
program.md		program.md
quant.h		quant.h
requirements.txt		requirements.txt
score.sh		score.sh

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Repository files navigation

The single-header C engine for KV-compressed LLM inference

The Problem

The Result

Llama 3.2 3B Instruct — PPL × Speed (FP32 KV NEON = 13.56 PPL @ 14.8 tok/s)

Cross-size validation (3 Llama-family models)

Context length gains (`turbo_kv_4b` + `q4` value cache)

Why quant.cpp?

Get Started in 60 Seconds

See It In Action: Book-in-a-Chat

How It Compares

vs llama.cpp: Quality at same bit budget

vs other TurboQuant implementations

vs production inference engines

Supported Models

KV Cache Compression

The Idea

Quality vs Compression

Modes

QK-norm Aware (Gemma 4)

Advanced Usage

Single-Header Mode

Browser Demo (WASM)

Docker & Server

Python

Backends & Performance

FAQ

Documentation

References & Citations

About

Uh oh!

Releases 8

Packages

Uh oh!

Contributors

Uh oh!

Languages

Folders and files

Latest commit

History

Repository files navigation

The single-header C engine for KV-compressed LLM inference

The Problem

The Result

Llama 3.2 3B Instruct — PPL × Speed (FP32 KV NEON = 13.56 PPL @ 14.8 tok/s)

Cross-size validation (3 Llama-family models)

Context length gains (turbo_kv_4b + q4 value cache)

Why quant.cpp?

Get Started in 60 Seconds

See It In Action: Book-in-a-Chat

How It Compares

vs llama.cpp: Quality at same bit budget

vs other TurboQuant implementations

vs production inference engines

Supported Models

KV Cache Compression

The Idea

Quality vs Compression

Modes

QK-norm Aware (Gemma 4)

Advanced Usage

Single-Header Mode

Browser Demo (WASM)

Docker & Server

Python

Backends & Performance

FAQ

Documentation

References & Citations

About

Topics

Resources

License

Code of conduct

Contributing

Security policy

Uh oh!

Stars

Watchers

Forks

Releases 8

Packages 0

Uh oh!

Contributors

Uh oh!

Languages

Context length gains (`turbo_kv_4b` + `q4` value cache)

Packages