Beluga
Stability AI's 2023 Llama fine-tunes — non-commercial, abandoned
Overview
Stable Beluga 1 (LLaMA-65B) and 2 (Llama-2-70B) are Stability AI's July 2023 Orca-style instruction fine-tunes; Beluga 2 briefly led the open-model leaderboard. Non-commercial license, no updates since, and Stability has exited language models entirely.
Pricing
Pricing shown for reference only. These figures reflect RECATOOLS research as of 11 Jul 2026 and may be out of date or incomplete. This is not financial or purchasing advice — always confirm the current price on the provider’s official website before making any decision.
Use cases
ASEAN Perspective
Beluga in Southeast Asia
ASEAN-region availability and pricing notes coming soon. Drop the editorial team a note via /contact/ if you can supply local context (Singapore/Malaysia/Indonesia/Thailand/Vietnam).
Stable Beluga is a closed chapter. The Llama-2-70B Beluga 2 briefly led the Hugging Face Open LLM Leaderboard in mid-2023 on the strength of Orca-style synthetic training data, but both models shipped under a non-commercial community license — free for research only — and Stability AI subsequently exited language models altogether to concentrate on image, video and audio generation. There is no API, no support and no successor; the weights remain downloadable on Hugging Face and Ollama. Worth knowing as an early proof that explanation-tuned synthetic data works, and as a lineage reference for open instruction tuning. Anyone choosing a 70B-class model in 2026 should look at Llama 3.3, Qwen or DeepSeek instead.
What people say
For a few weeks in mid-2023, Stability AI had the top-ranked open language model in the world. Stable Beluga 2 — a Llama-2-70B fine-tuned on an Orca-style synthetic reasoning dataset — sat at the top of the Hugging Face Open LLM Leaderboard, with the LLaMA-65B-based Beluga 1 close behind. It was the company's proof that it could do more than Stable Diffusion.
That proof led nowhere. Both models shipped under a non-commercial community license, so nobody could build a business on them, and Stability's 2024 crisis — founder Emad Mostaque's exit, layoffs, emergency funding — ended the language-model programme outright. Under CEO Prem Akkaraju (since June 2024) the company recovered by narrowing to what it's good at: image, video and audio generation, with Stable Audio 3.0 the latest release in May 2026. StableLM and Beluga get no development, no support and no successors; the weights simply sit on Hugging Face and Ollama for anyone curious.
What's left is a decent case study. Beluga demonstrated early that Orca-style explanation-tuned synthetic data could push a Llama base toward GPT-3.5 territory, months before that became conventional wisdom, and quantised community builds kept it alive on home hardware through 2023. If you're tracing the lineage of open instruction tuning, it belongs in the footnotes. If you're picking a 70B-class model to run today, Llama 3.3, Qwen or DeepSeek distillations beat it on every axis and come with licenses you can actually use. The 4.3 score reads generous for a product, fair for a citation.
Summary of public user & expert reviews, compiled by RECATOOLS.
Notable facts
- StableBeluga 2 was the first 70B open model to break 90% on the MMLU benchmark, a comprehensive multi-subject knowledge test.
- Stability AI released Beluga models at a time when the company was primarily known for image generation, surprising many observers with its LLM research output.
- The System Prompt Tuning technique developed for Beluga became widely adopted for building role-specific AI assistants that reliably maintain personas.
Frequently asked questions
About this listing
This entry was compiled from publicly available data including Beluga's official website, press releases, documentation, and reputable third-party publications. RECATOOLS is not affiliated with Beluga unless explicitly stated.
Third-party AI tools update their pricing, features, availability, and policies frequently. Information here may be outdated by the time you read this — we make reasonable efforts to keep listings current, but cannot guarantee absolute accuracy.
For the latest details, please refer to Beluga directly →
Spotted something out of date? Suggest an update →
Beluga in the news
Alternatives to Beluga
More in LLMs & Chat
