Dolly
Open-source instruction-following LLM that fine-tunes on a single GPU
Dolly is a great teaching tool for LLM fine-tuning on a single GPU, with a permissive Apache 2.0 license. But it's archived, and its weak math and coding performance means it's not for production. Use it to learn, then graduate to Llama 3 or Mistral.
Verified 5d ago · liveness 66/100 · cite: rightaichoice.com/tools/dolly
- Researchers experimenting with LLM fine-tuning on limited compute
- Developers needing a commercially safe open-source LLM for demos and prototypes
- Hobbyists training their first instruction-following model
- Teams exploring lightweight alternatives to large proprietary models
- Production chatbots requiring low latency and high reliability
- Tasks needing strong math, code, or complex reasoning
- Users seeking ongoing updates or community support
We scan live Reddit threads, YouTube comments, X posts, G2 reviews and other communities — and hand you an honest verdict in under a minute.
- Honest verdict, not marketing
- Real pros & cons from real users
- Attributed quotes with receipts
3 free scans · no card needed
Skip Dolly if you need a model for production use, require strong math or code capabilities, or want a project with active maintenance and community support.
You'll need your own GPU for fine-tuning (A100 or similar) or pay for cloud compute, which can add hundreds of dollars per hour.
Dolly is free and open-source, so the main cost is your compute and time. Compared to commercial APIs like OpenAI's GPT-4 (usage-based) or other open models like Llama 3 (also free but more modern), Dolly fits best for learning and prototyping on a tight budget where you already have GPU access.
In short
Dolly — Open-source instruction-following LLM that fine-tunes on a single GPU. Best for Researchers experimenting with LLM fine-tuning on limited compute, Developers needing a commercially safe open-source LLM for demos and prototypes, Hobbyists training their first instruction-following model. Free to use.
What people actually say about Dolly — is it worth it?
We ran a structured research pass across product reviews, community discussions, and post-purchase forum threads to surface the patterns vendors won't publish themselves. Below: the recurring strengths, the hidden costs people mention most, and the cohort that consistently regrets adopting this tool.
50 mentions across 3 sources (Hacker News, GitHub, Lemmy) · researched Jul 3, 2026.
- +Open-source and Apache 2.0 licensed for commercial use.
- +Fine-tunes in ~30 minutes on a single machine.
- +Dolly-15k dataset is high-quality and community-driven.
- +Based on Pythia-12B, a well-known architecture.
- +Inference possible on consumer hardware with optimizations.
- −CUDA out-of-memory errors plague even large GPU instances.
- −Deepspeed setup is broken with missing shared library errors.
- −Model loading fails with standard transformers classes.
- −Documentation lacks key parameters for inference success.
- −Project activity has stalled with unresolved issues.
- • Requires expensive GPU hardware for training (e.g., p3.16xlarge at $5+/hr)
- • Potential cloud compute costs if not self-hosted
Viability Score
How well maintained and how widely used is Dolly? Built from what the vendor actually publishes (docs, changelog, tutorials, integrations, pricing), whether the site is live, and how much real users discuss it. How we calculate this
Last calculated: September 2026
How we score →Key Features
- Fine-tune on ~15k instruction records
- Based on Pythia-12B architecture
- Apache 2.0 license for commercial use
- Covers 7 instruction domains (brainstorming, classification, etc.)
- Inference via Hugging Face Transformers pipeline
- Weights on Hugging Face (databricks/dolly-v2-12b)
- Open-source training code on GitHub
- Fine-tune in ~30 minutes on a single A100 GPU
- Inference on A10 GPUs with 8-bit quantization
- Inference on V100 GPUs with float16
- Dataset released under CC-BY-SA
- Pre-trained from The Pile corpus
- Archived project (no updates since Oct 2023)
About Dolly
Dolly is Databricks' open-source instruction-following large language model, designed for developers and researchers who want hands-on LLM fine-tuning without a massive compute cluster. Based on EleutherAI's Pythia-12B, it's fine-tuned on ~15,000 instruction/response pairs generated by Databricks employees, covering seven capability domains from the InstructGPT paper: brainstorming, classification, closed QA, generation, information extraction, open QA, and summarization. The model is available on Hugging Face as databricks/dolly-v2-12b, with training code and weights public on GitHub under Apache 2.0, so you can inspect, modify, and commercially deploy it. Dolly's standout trait is efficiency: fine-tune the 12B model in about 30 minutes on a single A100 GPU, and run inference on A10 GPUs with 8-bit quantization or V100s with float16. The training dataset, databricks-dolly-15k, is released under CC-BY-SA, letting you reuse it for your own experiments. This makes Dolly a practical entry point for hobbyists, students, and small teams exploring instruction tuning without a big hardware budget. But don't mistake Dolly for a frontier model. Databricks openly admits it struggles with programming, math, complex reasoning, and factual accuracy — and it lacks some capabilities like well-formatted letter writing. The project is archived as of October 2023, so there are no updates or community support coming. You get the code, weights, and dataset, but no roadmap. Compared to modern open-weight models like Llama 3 or Mistral, Dolly is a learning tool rather than a production asset. Its permissive licensing and modest hardware requirements make it a safe pick for demos, prototypes, and educational projects, but for serious products you'll want an actively maintained model that handles complex tasks better.
Behind the Verdict
Dolly fills a specific niche: affordable, hands-on instruction tuning. If you're a researcher or hobbyist who wants to understand how fine-tuning works without renting a cluster, Dolly is a solid starting point — you can train it on one A100 in about 30 minutes, and the 15k dataset is openly licensed for reuse. But here's the catch: Dolly is archived. The last commit was in 2023, and there's no community support or updates coming. That's fine for a learning project, but it's a red flag if you're building anything for production. You'll be stuck with a model that can't handle coding, math, or complex reasoning well. Where Dolly really shines is in demos and prototypes. The Apache 2.0 license means you can deploy it commercially without legal headaches, and the hardware requirements are modest — you can run inference on A10s with 8-bit quantization. That makes it a low-risk way to validate an idea before committing to a paid API or a beefier model. Compared to Llama 3 or Mistral, Dolly is several generations behind. Those models are actively developed, much stronger on code and reasoning, and often run locally too. If your goal is a working product, skip Dolly. If your goal is to learn the mechanics of instruction tuning, Dolly's transparency and simplicity give it an edge. One more caveat: the model has known bias and hallucination issues, inherited from The Pile and Wikipedia-based training data. Don't use it for anything requiring factual accuracy or unbiased outputs. For a learning tool, that's acceptable; for a customer-facing app, it's a dealbreaker. In practice, we'd recommend Dolly for a weekend experiment or a university course, not for a startup's MVP. It's a footnote in LLM history, but a useful one for education.
Researching Dolly? Get your full AI stack in 60 seconds.
Free, no signup — tell us your goal and get tools matched to your budget & existing stack.
Real-world workflow fit
Concrete scenarios for the personas Dolly actually fits — and what changes day-one when you adopt it.
You want to fine-tune a model on a small custom dataset to experiment with instruction tuning.
Outcome: You download the Dolly repo, prepare your data in the required format, and run train_dolly.py on a single A100 GPU, getting a fine-tuned model in about 30 minutes. You can then evaluate it on your tasks.
You want to build a simple chatbot for a personal project without spending money on API calls.
Outcome: You load the pre-trained Dolly model from Hugging Face using the Transformers pipeline and run inference on your local A10 GPU, creating a brainstorming assistant that runs offline.
Use Cases
- Fine-tune a custom instruction-following model on domain-specific data for internal tools
- Experiment with LLM alignment techniques using a small, fast-to-train base model
- Deploy a chatbot for brainstorming or closed QA on a private dataset
- Teach yourself modern LLM fine-tuning with a fully open-source pipeline
- Create a lightweight assistant for classification or information extraction tasks
- Run inference on a single A10 GPU for low-cost proof-of-concept
Models Under the Hood
as of 2026-09-02
Limitations
- Dolly is an open-source large language model fine-tuned on 15,000 instruction/response pairs.
- It is based on the Pythia-12B architecture and is intended for research and development.
- The model was trained in about 30 minutes on a single A100 GPU and is commercially licensed under Apache 2.0.
- The GitHub repository is archived as of October 2023, indicating it is no longer actively maintained.
- It struggles with programming problems, mathematical operations, complex reasoning, and may produce factual errors or hallucinations.
as of 2026-08-21
Verification history
We have re-verified Dolly 7 times since . Each pass re-reads the vendor's own pages and re-checks every listed field against that evidence; passes where nothing had changed are marked as such.
- — re-verified summary, description, our verdict, our analysis, pricing model, pricing tiers, features, integrations, who it suits, who should skip it
- — re-verified summary, description, our verdict, our analysis, pricing model, pricing tiers, features, integrations, who it suits, who should skip it
- — re-verified summary, description, our verdict, our analysis, pricing model, pricing tiers, features, integrations, who it suits, who should skip it
- — re-verified summary, description, our verdict, our analysis, pricing model, pricing tiers, features, integrations, who it suits, who should skip it
- — re-verified summary, description, our verdict, our analysis, pricing model, pricing tiers, features, integrations, who it suits, who should skip it
- — re-verified summary, description, our verdict, our analysis, pricing model, pricing tiers, features, integrations, who it suits, who should skip it
Showing the 6 most recent of 7 verification passes.
Free to cite with attribution — this page re-verifies continuously.
12-month cost
Project the real annual outlay, including the implied monthly cost when only an annual tier is published.
Vendor list price only. Add-on usage, seat overages, and contract minimums are surfaced under Hidden costs & gotchas.
Plans compared
For each published Dolly tier: who it actually fits, and what it adds vs. the previous tier. Cross-reference the cost calculator above for projected annual outlay.
Open Source
$0
Ideal for
Researchers and hobbyists who want to learn LLM fine-tuning on a budget, with access to a single GPU.
What this tier adds
This is the starting tier: free Apache 2.0 license, includes training code, weights, and the dolly-15k dataset.
Where the pricing makes sense
The company stage and team size where Dolly's pricing actually pencils out — and where peers do it cheaper.
Dolly is free and open-source, so the main cost is your compute and time. Compared to commercial APIs like OpenAI's GPT-4 (usage-based) or other open models like Llama 3 (also free but more modern), Dolly fits best for learning and prototyping on a tight budget where you already have GPU access.
Setup time & first value
How long it actually takes to get something useful out of Dolly — broken out by persona, not the marketing-page minute.
For a researcher with GPU access, you can go from cloning the repo to fine-tuning in about 1-2 hours, including setup. For a developer using the pre-trained model for inference, expect under 30 minutes to have a basic pipeline working.
Switching to or from Dolly
How to bring data in from common predecessors and how to get it back out — written for the switcher, not the buyer.
- →From Any LLM API: If you're currently using a paid API like OpenAI, you can switch to Dolly for prototyping by downloading the model and running it on your own GPU, saving on per-token costs.
- ↗To Llama 3: If you need better performance and active maintenance, you can start with Llama 3, which has a larger ecosystem and ongoing updates.
Resources & Guides
Tutorials & Learning
![ROBLOX - Dolly - [Full Walkthrough]](https://img.youtube.com/vi/w-9Nq6_94_I/mqdefault.jpg)
Official links
Tools that pair well with Dolly
Common stack mates teams adopt alongside Dolly, with the specific reason each pairing earns its keep.
Zhipu AI
Zhipu AI's GLM-5.2 delivers 1M lossless context, open-source SOTA coding, and autonomous agent APIs for enterprises.
StableLM
StableLM: open-source, self-hostable LLM suite for transparent text and code generation
Falcon LLM
Open-weight multilingual AI with hybrid Transformer-Mamba architecture from TII.
Featured Head-to-Head Comparisons
Dolly vs Surge Ai
Dolly is a free, lightweight open-source LLM you can fine-tune yourself in 30 minutes, perfect for experimentation or building custom bots. Surge AI, in contrast, is a premium enterprise platform that delivers expert human feedback for training frontier models—if you need rigorous RLHF or benchmarks like those used by Anthropic, Surge is the clear choice. Choose Dolly for low-cost tinkering and Surge for production-grade alignment.
Dolly vs Praktika
If you want to improve your spoken language fluency with real-time corrections, Praktika is the clear choice—it’s a polished mobile app with AI tutors. But if you need a lightweight, openly-licensed LLM to fine-tune for custom instruction tasks, Dolly is unbeatable for its price (free) and speed of training. These tools serve entirely different needs, so your pick depends on whether you’re a language learner or a developer.
Alternatives to Dolly
View allZhipu AI
Zhipu AI's GLM-5.2 delivers 1M lossless context, open-source SOTA coding, and autonomous agent APIs for enterprises.
StableLM
StableLM: open-source, self-hostable LLM suite for transparent text and code generation
Falcon LLM
Open-weight multilingual AI with hybrid Transformer-Mamba architecture from TII.
Frequently Asked Questions
Categories
Used Dolly? Help shape our editorial sentiment research.