> ## Documentation Index
> Fetch the complete documentation index at: https://docs.tensormesh.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Compatibility Matrix

> Models, serving engines, cache features, and recovery scenarios supported on the latest Tensormesh release — and the exact versions to deploy.

This page lists what the current Tensormesh release supports: the models it serves,
the serving engines it runs in front of, the cache and storage features it provides,
and the failures it recovers from — along with the exact versions to deploy.

Tensormesh is released and supported as a single matched set, so the versions below
are the ones to run together.

{/* Auto-generated by scripts/generate-compatibility-matrix.mjs — do not edit this block by hand. */}

## Compatibility matrix

Tensormesh is released as a matched set. Deploy a single row's versions together —
the cache engine, operator, and Helm chart are supported as a unit.

| Cache engine | Operator                   | Helm chart | Serving image                |          Status          | As of      |
| ------------ | -------------------------- | ---------- | ---------------------------- | :----------------------: | ---------- |
| `v0.4.5`     | `v0.1.2`                   | `0.4.5`    | `lmcache/vllm-openai:v0.4.5` |        ✅ Supported       | 2026-05-27 |
| `v0.5.0`     | `nightly-20260508-c7394b8` | `0.4.5`    | `lmcache/vllm-openai:v0.5.0` | 🚧 Upgrade in validation | 2026-06-24 |

*Current as of **2026-06-24**.*

<Warning>
  Deploy these versions together and pin them explicitly — never `latest`.
  Mixing versions across the cache engine, operator, and chart is not supported.
</Warning>

## What's supported

### Supported models

Models supported on the current Tensormesh release.

| Model                    | Hugging Face ID                             | Status      |
| ------------------------ | ------------------------------------------- | ----------- |
| DeepSeek-R1-Distill 1.5B | `deepseek-ai/DeepSeek-R1-Distill-Qwen-1.5B` | ✅ Supported |
| GLM-4 9B                 | `THUDM/glm-4-9b`                            | ✅ Supported |
| Llama 3.2 1B             | `meta-llama/Llama-3.2-1B`                   | ✅ Supported |
| Qwen3 0.6B               | `Qwen/Qwen3-0.6B`                           | ✅ Supported |

<Note>On the roadmap: Llama 3.1 8B, MiniMax.</Note>

### Serving engines

Serving engines supported in front of the Tensormesh cache.

| Serving engine | Notes                        | Status      |
| -------------- | ---------------------------- | ----------- |
| vLLM           | Bundled in the serving image | ✅ Supported |

<Note>On the roadmap: NVIDIA Dynamo, SGLang, TensorRT-LLM, vLLM via stack-operator.</Note>

### Cache features & storage

Cache and storage capabilities available on the current release.

| Capability                         | Type       | Status      |
| ---------------------------------- | ---------- | ----------- |
| L2 filesystem offload (fs\_native) | Filesystem | ✅ Supported |

<Note>On the roadmap: GPUDirect Storage (GDS), Peer-to-peer KV transfer.</Note>

### Resilience

Failures Tensormesh detects and recovers from automatically.

| Scenario                          | Status         |
| --------------------------------- | -------------- |
| Cache engine pod failure recovery | ⚠️ Known issue |

<Note>On the roadmap: Network partition recovery, Operator pod failure recovery.</Note>

### Performance

Performance characteristics of the current release.

| Benchmark                         | Hardware | Status      |
| --------------------------------- | -------- | ----------- |
| L1 cache-hit speedup (Qwen3 0.6B) | H200     | ✅ Supported |

## How to read this

* **✅ Supported** — works on the current release.
* **⚠️ Known issue** — a problem we're aware of and actively tracking on the current
  release.
* **On the roadmap** — planned, not yet available. Absence from this page doesn't
  mean something can't work, only that it isn't supported yet.

Always deploy the versions in **Supported versions** above as a set, and pin them
explicitly rather than using `latest`.

If you need to attach an existing vLLM workload to Tensormesh, see
[Modify an Existing Deployment](/operator/installation/existing-deployment).