FreeToken quantization: Setup Guide for MoE Inference - Architecture

FreeToken quantization: Setup Guide for MoE Inference

Learn how FreeToken handles quantized MoE models, memory tiers, hardware requirements, caching, and local inference setup in 2026.

2026-08-25
FreeToken Team
Quick Guide
  • FreeToken quantization mainly concerns serving supported low-precision checkpoints efficiently on local hardware.
  • MoE design lets only a small subset of model parameters activate for each token.
  • VRAM limits do not remove the need for sufficient system RAM to hold the complete model.
  • Best hardware fit currently means Linux, NVIDIA GPUs, CUDA 13, and a recent driver.
  • Main advantage appears when a large MoE model cannot fit entirely inside GPU memory.

FreeToken quantization and the core serving model

FreeToken is an edge-native inference engine, not a new language model and not a standalone quantization algorithm. In the context of FreeToken quantization, the important question is how the runtime serves supported low-precision checkpoints when the full mixture-of-experts model exceeds available VRAM.

A mixture-of-experts model stores a large pool of experts but routes each token through only a limited selection. DeepSeek-V4-Flash is described as having 284 billion total parameters, with approximately 13 billion active for each token. That sparse activation makes local execution more practical, while the complete expert pool still creates a major memory and transfer challenge.

ConceptWhat it meansWhy it matters
Quantized checkpointA model stored using reduced-precision weightsLowers storage and memory pressure
MoE modelA model with many experts and sparse token routingReduces active computation per token
Active parametersParameters used for the current tokenDetermines much of the immediate compute load
Full checkpointAll model and expert weightsStill must reside somewhere on the system
Edge servingInference across GPU, CPU, RAM, and PCIeMakes consumer hardware a unified runtime

The paper describes FreeToken supporting multiple model families and precision formats. Its evaluation includes a natively MXFP4-quantized DeepSeek-V4-Flash checkpoint, an NVFP4 release for an 8 GB laptop configuration, and Qwen3.6-35B-A3B in BF16. This means “quantization support” depends on the specific model checkpoint and runtime path rather than on a universal conversion rule.

Video Highlights:

  • FreeToken is designed for large MoE models that exceed GPU memory.
  • Adaptive expert caching keeps frequently routed experts in VRAM.
  • CPU execution and PCIe transfers can work together on cache misses.
  • Long coding-agent sessions are an important target workload.
Practical Interpretation

Choose a supported quantized checkpoint first, then calculate total system memory needs. An 8 GB GPU can accelerate a larger model, but it cannot store the entire checkpoint by itself.

How quantized MoE weights move through memory

FreeToken organizes inference around a two-level expert-memory hierarchy. The host system keeps the complete routed-expert pool, while non-expert weights remain on the GPU. Available VRAM is then used as an elastic expert cache shared across MoE layers.

This arrangement changes the role of quantization. Reduced-precision weights lower the size of the host-resident model and reduce transfer volume, but the runtime still has to decide which experts should be in VRAM, which missing experts should be transferred, and which can be executed directly by the CPU.

Memory layerPrimary contentsRuntime role
GPU memoryNon-expert weights, KV cache, selected expertsFast execution and active-state storage
Host RAMComplete expert poolSource of truth for model weights
PCIe linkExpert transfers and activation trafficConnects CPU-side storage to GPU execution
NVMe storageCheckpoint files and FTW dataSupplies model data during startup
CPU cache pathRecently used host-side dataSupports direct execution of misses

During prefill, FreeToken uses full-layer double buffering when enough cache capacity is available. While the GPU computes one layer, the next layer’s experts can stream over PCIe. This overlaps data movement with computation instead of exposing every transfer as separate GPU idle time.

During decode, routing becomes more fine-grained. The runtime maintains a shared LRU cache based on recently selected experts. A cache hit executes on the GPU. A miss can be added to the cache through PCIe or executed directly on the CPU, depending on measured host and transfer bandwidth.

The bandwidth-adaptive policy is central to this design. A fast PCIe connection may favor transferring more missing experts, while stronger host-memory bandwidth can make direct CPU execution more attractive. The decision is made for the deployed machine rather than copied from a fixed hardware profile.

Memory Warning

Quantization reduces checkpoint size but does not eliminate memory requirements. Large models still need enough RAM for the complete host-resident expert pool, plus operating-system and application overhead.

Supported formats, hardware, and performance profile

FreeToken’s documented accelerated setup is currently specialized. The command-line path described in the available material requires Linux on an x86-64 computer, an NVIDIA GPU, CUDA 13, and a recent driver. The project highlights RTX 30, RTX 40, and RTX 50 series hardware.

Hardware or platformStatus in the available documentationKey consideration
RTX 30 seriesHighlighted supportHost RAM remains important for oversized MoE models
RTX 40 seriesHighlighted supportPCIe and CPU bandwidth affect cache misses
RTX 50 seriesHighlighted supportStrong fit for large local MoE experiments
RTX 4060 laptopEvaluated configuration8 GB VRAM, 32 GB system memory, NVFP4 checkpoint
RTX PRO 6000 BlackwellFrontier-scale evaluationUsed for GLM-5.2 demonstration
Apple SiliconNo comparable path describedDo not assume a native equivalent
CPU-only executionNot the primary targetSpecialized GPU serving is the focus

The published evaluation reports 77–83 tokens per second for Qwen3.6-35B-A3B and 22–25 tokens per second for DeepSeek-V4-Flash on an RTX 5090 configuration. On an RTX 4060 laptop with 8 GB of VRAM and 32 GB of system memory, the official NVFP4 Qwen configuration reached 39.3 tokens per second.

A separate workstation result served GLM-5.2, described as a 753 billion parameter model, at 14.9 tokens per second on a single RTX PRO 6000. These figures are tied to specific checkpoints, hardware, workloads, and precision formats. They should be treated as reference points rather than universal performance guarantees.

Model or configurationPrecision or formatReported hardwareReported result
Qwen3.6-35B-A3BBF16RTX 509077–83 tokens per second
DeepSeek-V4-FlashMXFP4 routed expertsRTX 509022–25 tokens per second
Qwen3.6-35B-A3BOfficial NVFP4 releaseRTX 4060 laptop, 8 GB VRAM39.3 tokens per second
GLM-5.2NVFP4 routed expertsRTX PRO 6000 Blackwell14.9 tokens per second
Qwen3.6-35B-A3BCommunity-reported quantized testRTX 5080 systemAround 100 tokens per second reported

The strongest use case is a model that does not fit in VRAM but remains manageable in system memory. If a quantized model fits completely inside the GPU, a general-purpose runtime may already deliver excellent speed, and FreeToken’s transfer-oriented advantages may be less important.

Best-Fit Scenario

FreeToken is most compelling when an NVIDIA GPU, sufficient system RAM, and a large MoE checkpoint must work together for interactive local inference.

FreeToken quantization setup workflow

Use this workflow to evaluate a supported quantized model without treating published benchmark numbers as guaranteed results.

1

Confirm the platform

Verify that the machine uses Linux, an x86-64 processor, a supported NVIDIA GPU, CUDA 13, and a recent driver. Also record available VRAM, system RAM, PCIe link width, and host-memory bandwidth.

2

Select a supported checkpoint

Choose an official or documented checkpoint with a compatible precision format, such as MXFP4, NVFP4, or the listed BF16 evaluation configuration. Confirm that the model family is supported before downloading its weights.

3

Calculate total memory needs

Treat GPU memory as acceleration space, not as the complete storage location. Reserve enough system RAM for the full expert pool, runtime state, the operating system, and the growing KV cache.

4

Prepare the runtime format

Use the project’s documented loading path. Where applicable, the FreeToken Weight format stores expert banks in the runtime layout, reducing checkpoint discovery and repacking work during startup.

5

Test with your real workload

Measure prompt latency, time to first token, decode speed, and stability across multi-turn sessions. Coding agents and tool calls can reveal behavior that short single-prompt tests miss.

The runtime can dynamically resize and rebuild the GPU expert cache at scheduler safe points. This is useful when browser windows, desktop applications, or other GPU workloads change the available VRAM during a session.

For agentic workloads, context handling matters as much as raw decode speed. FreeToken anchors recurrent-state checkpoints at semantic boundaries such as thinking segments, tool calls, tool outputs, and conversation turns. When an agent edits a prior block, the runtime can reuse the unchanged prefix and re-prefill only the new suffix.

Test metricWhat to recordWhy it matters
VRAM usageCache, KV cache, and non-expert allocationShows whether memory is balanced
System RAM usageHost-resident model and runtime overheadDetects memory pressure
Time to first tokenAverage and slowest turnsReveals prefill and context costs
Decode speedTokens per second by workloadEnables fair engine comparisons
Cache behaviorHit and miss rateShows whether expert locality helps
Session stabilityLong context and repeated tool callsTests practical agent serving
Benchmarking Advice

Run the same model, precision, prompt sequence, and agent harness across engines. Compare long-session behavior separately from short single-turn throughput.

Readiness checklist and engine comparison

Before adopting FreeToken for regular local serving, check the constraints that most often determine whether the setup is practical.

Readiness Checklist:

  • Confirm Linux, x86-64, NVIDIA GPU, CUDA 13, and a recent driver
  • Select a documented model and compatible quantization format
  • Reserve system RAM for the complete host-resident checkpoint
  • Measure PCIe transfer and CPU memory bandwidth on the target machine
  • Test long-context and tool-calling workloads before daily use

FreeToken is not positioned as a universal replacement for llama.cpp. llama.cpp supports a broader range of operating systems, processors, GPU vendors, Apple Silicon devices, and model formats. FreeToken instead focuses on oversized MoE models and the coordination of GPU memory, CPU execution, host RAM, and PCIe transfers.

RuntimeMain strengthMain limitation for this use case
FreeTokenAdaptive MoE serving across GPU and CPU resourcesNarrower platform and model ecosystem
llama.cppBroad hardware and model compatibilityStatic hybrid placement can miss changing expert locality
KTransformersCPU expert execution and hybrid servingReported policies may be less adaptive to hardware balance
OllamaAccessible local model workflowNot the primary specialized target for oversized MoE serving

The project also provides OpenAI-compatible and Anthropic-compatible APIs, allowing local models to connect with supported coding and agent tools. Compatibility at the API layer does not guarantee identical behavior across every client, so test authentication, context handling, tool calls, and timeout settings individually.

For the technical design and evaluation details, read the FreeToken research paper. The paper identifies FreeToken as an Apache 2.0 open-source system and lists the release destination as flashml.ai.

Editor’s Recommendation

Start with one supported quantized MoE checkpoint and a reproducible benchmark. Expand only after confirming memory headroom, acceptable first-token latency, and stable agent sessions.

FreeToken quantization FAQ

Q: Is FreeToken itself a quantization algorithm?

No. FreeToken is an inference and serving engine. The term FreeToken quantization usually refers to running compatible low-precision checkpoints through its MoE serving system.

Q: Does an 8 GB GPU run a 35B model using only 8 GB of memory?

No. The evaluated laptop configuration used 8 GB of VRAM and 32 GB of system memory. The GPU handled acceleration while host memory held the remaining model weights.

Q: Which quantization formats are discussed for FreeToken?

The available evaluation describes MXFP4 routed experts for DeepSeek-V4-Flash, an official NVFP4 Qwen3.6 release for the laptop test, and BF16 for the main Qwen3.6 comparison.

Q: When is FreeToken more useful than a general local runtime?

It is most useful when a large MoE model exceeds VRAM, the computer has sufficient system RAM, and adaptive expert caching or CPU-GPU coordination can reduce transfer stalls.