Lamini

license

Proprietary

source

closed

engine

closed(hosted memory-tuning/fine-tuning service on Lamini/AMD servers)

client-sdk

open(Apache-2.0, lamini-ai/lamini)

data

closed

training-code

closed

access

API-key-gated

Proprietary managed fine-tuning platform; no source is published for the engine in any of the on-demand, reserved, or self-managed deployment tiers, and the only open artifact is a thin API client SDK, so the product itself remains closed.

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

  Repo title/description read "The Official Python Client for Lamini's API"; GitHub's own repo metadata records it public, not archived (isArchived:false), licensed Apache-2.0, with website field pointing to https://lamini.ai/. No fine-tuning or inference engine code lives in the repo -- only the API client.

• https://docs.lamini.ai/about/ recorded 2026-08-09

  States "Lamini Platform is available in three different deployment models": On-Demand (fully-managed training/inference at app.lamini.ai, pay-as-you-go), Reserved (dedicated GPUs on Lamini's own infrastructure, per-GPU pricing), and Self-Managed (run in the customer's own environment/GPUs -- on premise, VPC, or air-gapped -- per-GPU pricing). Every tier is a paid deployment of Lamini's own platform; none exposes a published repository for the engine.

• https://docs.lamini.ai/LICENSE/ recorded 2026-08-09

  Lists only third-party open-source notices bundled into the self-managed distribution -- vLLM (Apache-2.0), LlamaFactory (Apache-2.0, credited for supporting "MoME training"), litellm (MIT), and Ray (Apache-2.0 modifications) -- with no license grant for Lamini's own platform code anywhere on the page.

No independent user, usage or download figure is published for the platform. The team's own claim of having "shipped LLMs in production to over 1 billion users" describes the founders' prior careers rather than this product, and the hosted app returns an error, so nothing usable can be read from it either. No level is assigned rather than one inferred from press coverage.

• https://docs.lamini.ai/about/ recorded 2026-08-09

  The "Who are we?" section states the team "shipped LLMs in production to over 1 billion users" as a claim about the founders' collective prior careers (not the Lamini product's usage), and publishes no download, customer, or active-user count for the Lamini platform itself.

• https://app.lamini.ai/ recorded 2026-08-09

  The hosted On-Demand app currently returns HTTP 530 (a Cloudflare origin-unreachable error), so no live usage signal is observable from the product surface itself today.

The vendor markets memory-tuning with claims of roughly 95% factual accuracy and 32x compression, but those claims remain unverified: the arXiv paper behind them was withdrawn by its own authors, and the engine is closed, so none of it can be checked against the feature comparison used to score this category. No score is assigned rather than a fabricated one.

• https://arxiv.org/abs/2406.17642 recorded 2026-08-09

  The paper behind Lamini's Mixture of Memory Experts claims ("Banishing LLM Hallucinations Requires Rethinking Generalization") was withdrawn by its authors. Submission history: "[v2] Wed, 3 Sep 2025 08:09:31 UTC (1 KB) (withdrawn)", with the author's own comment "I want to revisit some of the experiments in this paper, specifically figure 5."

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

  The public repository is the API client only; no engine internals, training code, or benchmark harness is published to independently assess Lamini's memory-tuning claims against.