Skip to main content

kllm: free GPU LLMs for Claude Code, opencode and any OpenAI client

kllm: free GPUs for your coding agent. pipx install kllm-gateway. OpenAI and Anthropic API on localhost

kllm runs open-weight LLMs on Kaggle's free GPUs and gives your coding tools a local API for them. It serves Qwen and other GGUF models (including abliterated ones) through llama.cpp on two free Tesla T4s, behind an OpenAI and Anthropic compatible endpoint on 127.0.0.1. No local GPU, no API bill.

15,481 requests 98.95% succeeded 750M prompt tokens 89.8% from cache 395K largest context

Six weeks of one developer's Claude Code and opencode sessions (Oct 6 to Nov 16 2026), from kllm's request log.

Install and run

pipx install kllm-gateway      # or: uv tool install kllm-gateway, or: pip install kllm-gateway
kllm                           # opens the panel; paste your Kaggle token once
claude-kllm                    # Claude Code on a 35B model in Kaggle

The package is kllm-gateway; it installs the kllm and claude-kllm commands. You need a Kaggle account with a verified phone number (Kaggle's rule for GPU access). The first request starts a GPU in about 8 to 10 minutes; after that, requests go straight through.

kllm's first-run page: open your Kaggle settings, verify your phone, create a token, then paste it or drop kaggle.json

Models measured on two T4s

Model Best for Context Speed
qwen3.8-27b family Highest answer quality; vision variant 128K or 256K 16 to 19 tok/s over 6,889 real requests
qwen3.6-35b-abliterated (default) Long agent sessions at speed 256K 70 tok/s short, 43 tok/s at 209K
qwen3-coder-30b-abliterated Fastest turnaround 96K 84 tok/s short

Use it from any tool

Tool How
Claude Code claude-kllm
opencode, Codex CLI, Gemini CLI, Antigravity, Cursor the panel's Connect page writes their config
OpenAI clients base URL http://127.0.0.1:8080/v1, model kllm
Anthropic clients ANTHROPIC_BASE_URL=http://127.0.0.1:8080

When a Kaggle notebook nears its 12-hour limit, kllm starts a successor and moves each conversation with its prompt cache warm, so long sessions continue. Idle GPUs stop by themselves to save your weekly quota.

Use your own Kaggle account and follow Kaggle's Terms of Service. MIT licensed.

Metadata

Release files for kllm-gateway 0.3.4

For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.

Source distribution (sdist)

Source distribution for kllm-gateway 0.3.4
File Size Uploaded Provenance Actions
kllm_gateway-0.3.4.tar.gz 4.9 MB With provenance Details for kllm_gateway-0.3.4.tar.gz

Built distribution (wheel)

Table of built distributions (wheels) for kllm-gateway 0.3.4
File Interpreter ABI Platform Actions
kllm_gateway-0.3.4-py3-none-any.whl Python 3 none any Details for kllm_gateway-0.3.4-py3-none-any.whl

Total release size: 8.7 MB

Release files / kllm_gateway-0.3.4.tar.gz

Download URL kllm_gateway-0.3.4.tar.gz
Size 4.9 MB
Tags Source
SHA-256 checksum
How to use checksums
a1cc93ee137ecc8d7d668f5e1a543debae43d4036a5d502652dbcd9fb1996b9b
BLAKE2b-256 checksum
How to use checksums
1dac4280fd38098a7e52a49be4003d441afc3d17d3ebf8c243ce46cb81d7dfe9
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
Yes
Uploaded via twine/7.0.0 CPython/3.13.16

Provenance

Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.

PyPI Publish Attestation

PyPI verified that this artifact, at this checksum, originated from the publisher listed below.

Signed by GitHub Actions, verified by PyPI on Nov 19, 2026.

Transparency log

Release files / kllm_gateway-0.3.4-py3-none-any.whl

Download URL kllm_gateway-0.3.4-py3-none-any.whl
Size 3.8 MB
Tags Python 3
SHA-256 checksum
How to use checksums
6985a51f42baddf8070a537879b08eb2e8ac803c13ced1e66df0f00bf8a72077
BLAKE2b-256 checksum
How to use checksums
ba361744d1a0b70f5d632e765f1360dc70e338e0263400c11e2bf4c943c9baf9
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
Yes
Uploaded via twine/7.0.0 CPython/3.13.16

Provenance

Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.

PyPI Publish Attestation

PyPI verified that this artifact, at this checksum, originated from the publisher listed below.

Signed by GitHub Actions, verified by PyPI on Nov 19, 2026.

Transparency log

Security for kllm-gateway 0.3.4

Provenance

Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.

Full provenance

Both files in this release have been uploaded with provenance metadata.

Select from 2 files

Attestation for kllm_gateway-0.3.4.tar.gz

PyPI Publish Attestation

PyPI verified that this artifact, at this checksum, originated from the publisher listed below.

Signed by GitHub Actions, verified by PyPI on Nov 19, 2026.

Transparency log

A note on package security

PyPI is a repository for distributing packages, not a service for auditing them. While third-party security researchers and our staff do their best to detect and remove malicious packages, PyPI cannot guarantee the quality, security or suitability of any given package.

To learn more about protecting yourself from malicious actors, visit our package security documentation.

Anthropic, PBC Visionary sponsor Bloomberg Visionary sponsor Hudson River Trading Visionary sponsor Meta Visionary sponsor NVIDIA Visionary sponsor Microsoft Sustainability sponsor Depot Continuous Integration AWS Cloud computing and Security Sponsor Datadog Monitoring Fastly CDN Google Download Analytics Sentry Error logging StatusPage Status page