gpt2-large fine-tuned on CommonGen, used as the base language model for the CommonGen experiments in Mitigating Bias in Locally Constrained Decoding via Tractable Proposals (arXiv:2606.01926). This is a plain causal language model: it supplies the base distribution that GCD and P-GCD steer. The tractable proposal it is paired with is the HMM at which shares its 50257-token vocabulary. configs/common-gen.yaml in github.com/MhDang/gelatwo already points at this checkpoint, so the CommonGen runs need no override.
GPT-2 Large is the 774M parameter version of GPT-2, a transformer-based language model created and released by OpenAI. The model is a pretrained model on English language using a causal language modeling (CLM) objective.
Parameters812M
Context—
Weights16.2 GB
Licensemit
AccessOpen weights
Monthly Downloads1.2M
Runs On
What it takes to serve gpt2-large (812M parameters): the memory its weights need at each precision, and the cheapest way to rent enough data-center GPUs to hold them.
| Precision | Weights | Memory needed | Cheapest setup | Per hour | Also fits |
|---|---|---|---|---|---|
| 16-bit | 1.6 GB | 1.9 GB | 1x MI300X (192 GB) Vultr |
$1.85 | 1x H100 $1.99 · 1x MI325X $2.00 |
| 8-bit | 0.8 GB | 1.0 GB | 1x MI300X (192 GB) Vultr |
$1.85 | 1x H100 $1.99 · 1x MI325X $2.00 |
| 4-bit | 0.4 GB | 0.5 GB | 1x MI300X (192 GB) Vultr |
$1.85 | 1x H100 $1.99 · 1x MI325X $2.00 |
Memory is the weights at that precision plus 20% for the runtime and a short context; a long context needs more. Prices are the lowest on-demand hourly rates in the SAVRN Index, read Sep 18, 2026.
SAVRN's Notes on gpt2-large
Published March 2, 2022, this English causal language model moves 1,243,231 downloads a month. The configuration counts 812M parameters; the publisher says 774M. At 16-bit the weights are 1.6 GB and need 1.9 GB, 8-bit needs 1.0 GB and 4-bit 0.5 GB, so the cheapest host in our data, one MI300X with 192 GB at $1.85 per hour, is far more card than it needs. Prove a pipeline on it, then move the real workload in.
MIT is a short permissive license: keep the copyright and permission notice and you can use, modify and redistribute it commercially. Our facts record no context length for it, so confirm the window in the configuration you load. The repository totals 26.3 GB across 29 files because it ships six formats, and the one paper it cites, arXiv:1910.09700, covers the carbon emissions of machine learning, not this model.
Model Card
By OpenAI community, published under mit, revision 32b71b12589c.
Table of Contents
- Model Details
- How To Get Started With the Model
- Uses
- Risks, Limitations and Biases
- Training
- Evaluation
- Environmental Impact
- Technical Specifications
- Citation Information
- Model Card Authors
Model Details
Model Description: GPT-2 Large is the 774M parameter version of GPT-2, a transformer-based language model created and released by OpenAI. The model is a pretrained model on English language using a causal language modeling (CLM) objective.
- Developed by: OpenAI, see associated research paper and GitHub repo for model developers.
- Model Type: Transformer-based language model
- Language(s): English
- License: Modified MIT License
- Related Models: GPT-2, GPT-Medium and GPT-XL
- Resources for more information:
- Research Paper
- OpenAI Blog Post
- GitHub Repo
- OpenAI Model Card for GPT-2
- Test the full generation capabilities here: https://transformer.huggingface.co/doc/gpt2-large
How to Get Started with the Model
Configuration
- Architecture
- GPT2LMHeadModel
- Vocabulary size
- 50,257
- Model type
- gpt2
Identity and Version
- Repository
- openai-community/gpt2-large
- Publisher
- OpenAI community
- Task
- Text generation
- Modality
- Text
- Library
- transformers
- Parameters
- 812M parameters
- Languages
- en
- Revision
- 32b71b12589c2f8d625668d2335a01cac3249519
- First published
- 2022-03-02
- Last updated
- 2024-02-19
Files and Weights
29 files, 26.3 GB in total. The weights are 8 files totalling 16.2 GB in bin, h5, msgpack, onnx, ot, safetensors.
Weights8 files · 16.2 GB
Configuration6 files · 2.1 KB
Tokenizer8 files · 6.2 MB
Documentation1 file · 12.3 KB
Other5 files · 10.1 GB
Repository1 file · 789 B
Every file
| File | Type | Size | SHA-256 |
|---|---|---|---|
| flax_model.msgpack | Weights | 3.1 GB | 9aa10eb25a62 |
| model.safetensors | Weights | 3.2 GB | 5f47f3e12f91 |
| onnx/decoder_model.onnx | Weights | 2.7 MB | 6928b9ea9cef |
| onnx/decoder_model_merged.onnx | Weights | 5.3 MB | fb52fbf6a99a |
| onnx/decoder_with_past_model.onnx | Weights | 2.7 MB | 620cad159131 |
| pytorch_model.bin | Weights | 3.2 GB | 8d3ade6b55ac |
| rust_model.ot | Weights | 3.5 GB | 4da5e3c0a8b1 |
| tf_model.h5 | Weights | 3.1 GB | 71e91ce5f112 |
| config.json | Configuration | 666 B | — |
| generation_config.json | Configuration | 124 B | — |
| generation_config_for_text_generation.json | Configuration | 165 B | — |
| onnx/config.json | Configuration | 886 B | — |
| onnx/generation_config.json | Configuration | 119 B | — |
| onnx/special_tokens_map.json | Configuration | 99 B | — |
| README.md | Documentation | 12.3 KB | — |
| onnx/Constant_164_attr__value | Other | 1.0 MB | ae525b5c247f |
| onnx/Constant_174_attr__value | Other | 1.0 MB | ae525b5c247f |
| onnx/decoder_model.onnx_data | Other | 3.4 GB | 37e3f7ff14aa |
| onnx/decoder_model_merged.onnx_data | Other | 3.4 GB | 37e3f7ff14aa |
| onnx/decoder_with_past_model.onnx_data | Other | 3.4 GB | 37e3f7ff14aa |
| .gitattributes | Repository | 789 B | — |
| merges.txt | Tokenizer | 456.3 KB | — |
| onnx/merges.txt | Tokenizer | 456.3 KB | — |
| onnx/tokenizer.json | Tokenizer | 2.1 MB | — |
| onnx/tokenizer_config.json | Tokenizer | 234 B | — |
| onnx/vocab.json | Tokenizer | 798.2 KB | — |
| tokenizer.json | Tokenizer | 1.4 MB | — |
| tokenizer_config.json | Tokenizer | 26 B | — |
| vocab.json | Tokenizer | 1.0 MB | — |
License and Download
- License
- mit
- Access
- Open weights, no gate
- Download size
- 16.2 GB
Download from OpenAI community
Released by OpenAI community through its official repository on Hugging Face. Read the license.
Built From
- Described by arXiv:1910.09700
Memory Requirements
| Precision | Weights in memory |
|---|---|
| As published | 16.2 GB |
| 16-bit | 1.6 GB |
| 8-bit | 0.8 GB |
| 4-bit | 0.4 GB |
Weights only, from the published parameter count; the key-value cache and runtime add to this.
Built on This Model
- Derived fromgelatwo-common-gen-gpt2-large
Compare gpt2-large
Questions About gpt2-large
How much GPU memory does gpt2-large need?
About 1.9 GB at 16-bit and 0.5 GB at 4-bit: the weights (812M parameters) plus a working margin. A long context needs more.
What is the cheapest GPU to run gpt2-large on?
At 16-bit, 1x MI300X from $1.85 an hour; at 4-bit, 1x MI300X from $1.85 an hour, at the lowest on-demand prices the SAVRN Index lists.
Can I use gpt2-large commercially?
Yes. gpt2-large is released under MIT License. The MIT License is a short permissive license. It permits commercial use, modification and redistribution, provided the copyright notice and permission notice are included.
Similar Models
A masked diffusion language model adapted from Qwen3.5-0.8B. The backbone is hybrid: only its attention layers are made bidirectional, and the Gated DeltaNet layers stay causal. This is a base model, with no instruction tuning. Paper: dQwen3.5: Hybrid-Attention Diffusion Language Models. Code: https://github.com/AntonXue/dQwen Needs a CUDA GPU and transformers>=5.13 (tested with torch 2.7.1+cu128, flash-linear-attention 0.5.1). generate decodes the whole canvas at once, committing positions above a confidence threshold (tau=0.9); pass blocklength=32 for left-to-right block decoding, or tau=None, stepsperblock=k for a fixed budget. The 50B-token checkpoint from the paper is…
Qwen3 is the latest generation of large language models in Qwen series, offering a comprehensive suite of dense and mixture-of-experts (MoE) models. Built upon extensive training, Qwen3 delivers groundbreaking advancements in reasoning, instruction-following, agent capabilities, and multilingual support, with the following key features: - Uniquely support of seamless switching between thinking mode (for complex logical reasoning, math, and coding) and non-thinking mode (for efficient, general-purpose dialogue) within single model, ensuring optimal performance across various scenarios. - Significantly enhancement in its reasoning capabilities, surpassing previous QwQ (in thinking mode) and…
A 0.6B parameter model built in two stages: knowledge distillation from a 30B Thinking teacher to establish a structured reasoning backbone, then supervised fine-tuning on legal instruction data. 50x compression. Under 500MB quantized. Runs on a phone. The training order is the thesis: teach the model how to reason first (distillation from Thinking teacher), then teach it what to reason about (legal SFT). The Thinking teacher's extended deliberation traces transfer deeper reasoning structure than an Instruct teacher — critical when the student has only 0.6B parameters to work with. Qwen3-0.6B distilled from Qwen3-30B-A3B-Thinking-2507 — a Mixture-of-Experts model with 30B total parameters…
A 0.6B parameter model distilled from Qwen3-30B-A3B-Thinking on 6,122 STEM chain-of-thought samples. 50x parameter compression. The Thinking variant teacher produces richer extended reasoning traces than the Instruct variant, transferring deeper deliberation structure into the smallest possible student. The result: a model under 500MB quantized that produces structured STEM derivations because a 30B thinking model showed it how to reason. Two key differences from standard small-model distillation: 1. Thinking teacher, not Instruct teacher. The Qwen3-30B-A3B-Thinking variant generates extended internal reasoning before committing to an answer. Its softmax distributions are higher-entropy…
This is the model card of a transformers model that has been pushed on the Hub. Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. Use the code below to get started with the model. Carbon emissions can be estimated using the Machine Learning Impact calculator presented in Lacoste et al. (2019). This model is part of the Convergent Intelligence LLC: Research Division portfolio. All models in this portfolio are developed under the Discrepancy Calculus (DISC) framework — a measure-theoretic approach to understanding and controlling the gap between what a model should produce and what it actually produces. DISC treats training…