Axolotl

license

Apache-2.0(OSI)

source

public(axolotl-ai-cloud/axolotl)

core-gated

ungated(Apache-2.0 across the repo, no ee/ or enterprise directory, no license key and no closed package

Apache-2.0 across the whole repository, with the full source public and no proprietary managed training service, only deployment templates for third-party clouds. Nothing is withheld from the open package: there is no enterprise directory, no license-key gate and no closed component the open code depends on. The only paid offering is dedicated support by email.

• https://raw.githubusercontent.com/axolotl-ai-cloud/axolotl/main/LICENSE recorded 2026-08-11

  The body is the standard, unmodified Apache License, Version 2.0 text, with no custom copyright line or added restriction layered on top.

• https://github.com/axolotl-ai-cloud/axolotl recorded 2026-08-11

  Repo page for Axolotl, licensed Apache-2.0; the README states "This project is licensed under the Apache 2.0 License." Top-level directories are .github, .runpod, .vscode, cicd, deepspeed_configs, devtools, docker, docs, examples, image, scripts, src/axolotl, tests -- no ee/ or enterprise directory, no closed dependency and no license key. The only commercial line is "Need dedicated support? Please contact wing@axolotl.ai for options".

• https://raw.githubusercontent.com/axolotl-ai-cloud/axolotl/main/README.md recorded 2026-08-09

  Documents pip and Docker installation for self-hosting (uv pip install axolotl; docker run axolotlai/axolotl:main-latest) and lists third-party cloud deployment templates (RunPod, Vast.ai, PRIME Intellect, Modal, Novita, JarvisLabs, Latitude.sh) rather than a first-party managed or paid tier -- corroborates core-gated:ungated, which is not currently recorded as a controlled token on this product.

• https://axolotl.ai/ recorded 2026-08-09

  The marketing homepage frames the product as free and open source, self-hostable in your own cloud or Docker/Kubernetes setup, names third-party sponsors (e.g. Modal) rather than a first-party hosted offering, and carries no pricing page -- further corroborates core-gated:ungated.

~22.3K PyPI downloads in the trailing month, alongside ~12K GitHub stars. A popular config-driven fine-tuning tool in the open-model community, but download volume is modest next to TRL, PEFT and Unsloth, which puts reach in the 10K-100K band. The star count corroborates only; the level rests on downloads.

• https://pypistats.org/packages/axolotl recorded 2026-08-09

  Downloads last month: 12,389

• https://github.com/axolotl-ai-cloud/axolotl recorded 2026-08-09

  stargazerCount 12330 (corroborating signal, not the adoption basis)

• https://pypi.org/pypi/axolotl/json recorded 2026-08-09

  info.name = axolotl, info.version = 0.18.0: a live PyPI project named axolotl exists, even though the product file's artifact list carries no pypi entry and the warehouse signal_pypi table has no row for it (see findings).

Very wide coverage of both methods and parallelism: full fine-tuning, LoRA, QLoRA, GPTQ, quantization-aware training, the DPO/IPO/KTO/ORPO preference family, GRPO/GDPO and reward modeling, run over FSDP1 and FSDP2, DeepSpeed, multi-node Torchrun or Ray, and combined context, tensor and data parallelism, at BF16 or 4- to 8-bit precision. That is the broadest single-tool matrix among its peers here, and it trails Megatron-LM only on demonstrated extreme scale.

• https://raw.githubusercontent.com/axolotl-ai-cloud/axolotl/main/README.md recorded 2026-08-13

  Feature/parallelism matrix: full fine-tuning, LoRA, QLoRA, GPTQ, QAT (int8/int4/FP8/NVFP4/MXFP4), DPO/IPO/KTO/ORPO, GRPO/GDPO, reward and process-reward modelling; FSDP1/FSDP2, DeepSpeed, multi-node Torchrun/Ray, Sequence Parallelism, plus newly added Expert Parallelism (EP) and Context Parallelism for hybrid SSM models -- matches or exceeds what capability.value already records.