Introduce ginterpreter worker

docs/content/design.md

@@ -124,6 +124,23 @@ Choose a worker type based on your application's needs.
     gunicorn myapp:app -k tornado
     ```
 
+=== "Interpreter (Experimental)"
+
+    !!! warning "Experimental — requires Python 3.14+"
+
+    **Sub-interpreter** worker using `InterpreterPoolExecutor`. Each request runs
+    in its own sub-interpreter with an independent GIL, enabling true CPU
+    parallelism without forking extra processes.
+
+    - True parallelism for CPU-bound workloads
+    - No keep-alive support
+    - Hooks (`ssl_context`, `pre_request`, `post_request`) are not supported
+    - See the [Interpreter Worker](ginterpreter.md) guide for details
+
+    ```bash
+    gunicorn myapp:app -k ginterpreter --threads 4
+    ```
+
 ## Comparison
 
 | Worker | Concurrency Model | Keep-Alive | Best For |
@@ -134,13 +151,15 @@ Choose a worker type based on your application's needs.
 | `gevent` | Greenlets | ✅ | I/O-bound, WebSockets, streaming |
 | `eventlet` | Greenlets | ✅ | **Deprecated** - use `gevent` instead |
 | `tornado` | Tornado IOLoop | ✅ | Native Tornado applications |
+| `ginterpreter` *(experimental)* | Sub-interpreters | ❌ | CPU-bound apps on Python 3.14+ |
 
 !!! tip "Quick Decision Guide"
 
     - **Simple app behind nginx?** → `sync` (default)
     - **Need keep-alive or moderate concurrency?** → `gthread`
     - **WebSockets, streaming, long-polling?** → `gevent` or ASGI worker
     - **FastAPI, Starlette, or async framework?** → ASGI worker
+    - **CPU-bound and on Python 3.14+?** → `ginterpreter` *(experimental)*
 
 ## When to Use Async Workers
 

docs/content/ginterpreter.md

@@ -0,0 +1,51 @@
+# Interpreter Worker
+
+!!! warning "Experimental"
+    The `ginterpreter` worker is experimental and requires Python 3.14+. The API
+    and behavior may change in future releases.
+
+The interpreter worker uses Python's `InterpreterPoolExecutor` to handle each
+request in a separate sub-interpreter. Each sub-interpreter runs in its own
+thread with an independent GIL, enabling true CPU parallelism without multiple
+processes.
+
+## Quick Start
+
+```bash
+gunicorn myapp:app --worker-class ginterpreter --threads 4
+```
+
+Or in a configuration file:
+
+```python
+# gunicorn.conf.py
+worker_class = "ginterpreter"
+threads = 4
+```
+
+## Configuration
+
+The interpreter worker uses the standard gunicorn settings. The most relevant ones:
+
+| Setting | Default | Description |
+|---------|---------|-------------|
+| `threads` | `1` | Number of sub-interpreters (i.e. concurrent requests per worker) |
+| `workers` | `1` | Number of worker processes |
+| `timeout` | `30` | Request timeout in seconds |
+| `graceful_timeout` | `30` | Time to wait for in-flight requests on shutdown |
+
+## Known Limitations
+
+The following features are **not supported**:
+
+- **`ssl_context` hook** — SSL contexts cannot be shared across sub-interpreters. Built-in SSL via `certfile`/`keyfile` works normally.
+- **`pre_request` / `post_request` hooks** — Callables cannot be passed to sub-interpreters.
+- **Keepalive connections** — Each connection is closed after the response.
+- **HTTP/2**
+- **Sendfile**
+- **`max_requests` / `max_requests_jitter`**
+
+## See Also
+
+- [Settings Reference](reference/settings.md) - All available settings
+- [Design](design.md) - Worker architecture overview

gunicorn/workers/__init__.py

@@ -12,4 +12,5 @@
     "tornado": "gunicorn.workers.gtornado.TornadoWorker",
     "gthread": "gunicorn.workers.gthread.ThreadWorker",
     "asgi": "gunicorn.workers.gasgi.ASGIWorker",
+    "ginterpreter": "gunicorn.workers.ginterpreter.InterpreterWorker",
 }