- Size
- 20.7B parameterscounted from the safetensors header on the Hub
- Licence
- unknown, Apache License 2.0Our reading, check the licence: You can use, change and redistribute the model, including commercially. Keep the licence and any notices, and say which files you changed; you do not have to publish your changes. It includes a patent licence that ends for anyone who sues over patents in the work.
- On the Hub since
- 7 Apr 2022the date the repo was created
- Context length
- 2k tokensfrom the model's config
- Kind
- Base model (no base model declared on the Hub)GPTNeoXForCausalLM
- On Hugging Face
- EleutherAI/gpt-neox-20b278k downloads in the last 30 days, 587 likes
Scores
- default setting41.1%XGen-7B Technical Report
Epoch's name: gpt-neox-20b. Best closed model on this set: gpt-3.5-turbo-1106, 87.4% (OpenAI). All 134 results
- Shots: 064.9%XGen-7B Technical Report
Epoch's name: gpt-neox-20b. Best closed model on this set: Inflection-1, 89.7% (Inflection AI). All 136 results
- Shots: 070.5%XGen-7B Technical Report
Epoch's name: gpt-neox-20b. Best closed model on this set: gpt-4-0314, 95.3% (OpenAI). All 113 results
- Shots: 038.8%XGen-7B Technical Report
Epoch's name: gpt-neox-20b. Best closed model on this set: gpt-3.5-turbo-1106, 86.0% (OpenAI). All 70 results
- Shots: 076.7%XGen-7B Technical Report
Epoch's name: gpt-neox-20b. Best closed model on this set: gpt-4o-mini-2024-07-18, 88.7% (OpenAI). All 113 results
- Shots: 066.1%XGen-7B Technical Report
Epoch's name: gpt-neox-20b. Best closed model on this set: claude-3-opus-20240229, 88.5% (Anthropic). All 137 results
Claimed by the maker Claimed
The model card has no claimed scores in a machine-readable form (the card's model-index), so none are shown. The maker may give scores in the card text: read the card.
Other views, which we link to and do not copy: look this model up on Artificial Analysis and the Arena leaderboard.
Independent scores from Epoch AI (CC-BY 4.0), fetched 19 Nov 2026: Epoch AI, ‘Capabilities & benchmarking’. Published online at epoch.ai. Retrieved from ‘https://epoch.ai/benchmarks’ [online resource]. Each row links its own source. What the badges mean.
Run it yourself
Memory needed = weights + context cache + 0.50 GB for the runtime. Context cache at 8k tokens: 8.3 GB, worked out from the model's config (EleutherAI/gpt-neox-20b), at 16-bit precision. Choose a machine above to see a verdict for each format. Method
GGUF for llama.cpp, Ollama, LM Studio and others
mradermacher/gpt-neox-20b-i1-GGUF (converted by mradermacher), 1.1k downloads. Sizes are the real files on the Hub, all parts added up.
| Format | Size | Memory needed | Get it |
|---|---|---|---|
| IQ3_XXS | 7.5 GBfile size | 16.3 GB7.5 + 8.3 + 0.50 | Commandsllama.cpp llama-server -hf mradermacher/gpt-neox-20b-i1-GGUF:IQ3_XXSOllama ollama run hf.co/mradermacher/gpt-neox-20b-i1-GGUF:IQ3_XXSDownload hf download mradermacher/gpt-neox-20b-i1-GGUF --include "gpt-neox-20b.i1-IQ3_XXS.gguf" --local-dir gpt-neox-20b-i1-GGUF |
| IQ3_XS | 8.1 GBfile size | 16.9 GB8.1 + 8.3 + 0.50 | Commandsllama.cpp llama-server -hf mradermacher/gpt-neox-20b-i1-GGUF:IQ3_XSOllama ollama run hf.co/mradermacher/gpt-neox-20b-i1-GGUF:IQ3_XSDownload hf download mradermacher/gpt-neox-20b-i1-GGUF --include "gpt-neox-20b.i1-IQ3_XS.gguf" --local-dir gpt-neox-20b-i1-GGUF |
| IQ3_S | 8.4 GBfile size | 17.1 GB8.4 + 8.3 + 0.50 | Commandsllama.cpp llama-server -hf mradermacher/gpt-neox-20b-i1-GGUF:IQ3_SOllama ollama run hf.co/mradermacher/gpt-neox-20b-i1-GGUF:IQ3_SDownload hf download mradermacher/gpt-neox-20b-i1-GGUF --include "gpt-neox-20b.i1-IQ3_S.gguf" --local-dir gpt-neox-20b-i1-GGUF |
| Q3_K_S | 8.4 GBfile size | 17.1 GB8.4 + 8.3 + 0.50 | Commandsllama.cpp llama-server -hf mradermacher/gpt-neox-20b-i1-GGUF:Q3_K_SOllama ollama run hf.co/mradermacher/gpt-neox-20b-i1-GGUF:Q3_K_SDownload hf download mradermacher/gpt-neox-20b-i1-GGUF --include "gpt-neox-20b.i1-Q3_K_S.gguf" --local-dir gpt-neox-20b-i1-GGUF |
| IQ3_M | 9.3 GBfile size | 18.0 GB9.3 + 8.3 + 0.50 | Commandsllama.cpp llama-server -hf mradermacher/gpt-neox-20b-i1-GGUF:IQ3_MOllama ollama run hf.co/mradermacher/gpt-neox-20b-i1-GGUF:IQ3_MDownload hf download mradermacher/gpt-neox-20b-i1-GGUF --include "gpt-neox-20b.i1-IQ3_M.gguf" --local-dir gpt-neox-20b-i1-GGUF |
| Q3_K_M | 10.0 GBfile size | 18.8 GB10.0 + 8.3 + 0.50 | Commandsllama.cpp llama-server -hf mradermacher/gpt-neox-20b-i1-GGUF:Q3_K_MOllama ollama run hf.co/mradermacher/gpt-neox-20b-i1-GGUF:Q3_K_MDownload hf download mradermacher/gpt-neox-20b-i1-GGUF --include "gpt-neox-20b.i1-Q3_K_M.gguf" --local-dir gpt-neox-20b-i1-GGUF |
| IQ4_XS | 10.3 GBfile size | 19.0 GB10.3 + 8.3 + 0.50 | Commandsllama.cpp llama-server -hf mradermacher/gpt-neox-20b-i1-GGUF:IQ4_XSOllama ollama run hf.co/mradermacher/gpt-neox-20b-i1-GGUF:IQ4_XSDownload hf download mradermacher/gpt-neox-20b-i1-GGUF --include "gpt-neox-20b.i1-IQ4_XS.gguf" --local-dir gpt-neox-20b-i1-GGUF |
| Q4_0 | 10.9 GBfile size | 19.6 GB10.9 + 8.3 + 0.50 | Commandsllama.cpp llama-server -hf mradermacher/gpt-neox-20b-i1-GGUF:Q4_0Ollama ollama run hf.co/mradermacher/gpt-neox-20b-i1-GGUF:Q4_0Download hf download mradermacher/gpt-neox-20b-i1-GGUF --include "gpt-neox-20b.i1-Q4_0.gguf" --local-dir gpt-neox-20b-i1-GGUF |
| Q4_K_S | 10.9 GBfile size | 19.7 GB10.9 + 8.3 + 0.50 | Commandsllama.cpp llama-server -hf mradermacher/gpt-neox-20b-i1-GGUF:Q4_K_SOllama ollama run hf.co/mradermacher/gpt-neox-20b-i1-GGUF:Q4_K_SDownload hf download mradermacher/gpt-neox-20b-i1-GGUF --include "gpt-neox-20b.i1-Q4_K_S.gguf" --local-dir gpt-neox-20b-i1-GGUF |
| Q3_K_L | 11.0 GBfile size | 19.7 GB11.0 + 8.3 + 0.50 | Commandsllama.cpp llama-server -hf mradermacher/gpt-neox-20b-i1-GGUF:Q3_K_LOllama ollama run hf.co/mradermacher/gpt-neox-20b-i1-GGUF:Q3_K_LDownload hf download mradermacher/gpt-neox-20b-i1-GGUF --include "gpt-neox-20b.i1-Q3_K_L.gguf" --local-dir gpt-neox-20b-i1-GGUF |
| Q4_K_M | 12.2 GBfile size | 21.0 GB12.2 + 8.3 + 0.50 | Commandsllama.cpp llama-server -hf mradermacher/gpt-neox-20b-i1-GGUF:Q4_K_MOllama ollama run hf.co/mradermacher/gpt-neox-20b-i1-GGUF:Q4_K_MDownload hf download mradermacher/gpt-neox-20b-i1-GGUF --include "gpt-neox-20b.i1-Q4_K_M.gguf" --local-dir gpt-neox-20b-i1-GGUF |
| Q5_K_S | 13.2 GBfile size | 22.0 GB13.2 + 8.3 + 0.50 | Commandsllama.cpp llama-server -hf mradermacher/gpt-neox-20b-i1-GGUF:Q5_K_SOllama ollama run hf.co/mradermacher/gpt-neox-20b-i1-GGUF:Q5_K_SDownload hf download mradermacher/gpt-neox-20b-i1-GGUF --include "gpt-neox-20b.i1-Q5_K_S.gguf" --local-dir gpt-neox-20b-i1-GGUF |
| Q5_K_M | 14.2 GBfile size | 23.0 GB14.2 + 8.3 + 0.50 | Commandsllama.cpp llama-server -hf mradermacher/gpt-neox-20b-i1-GGUF:Q5_K_MOllama ollama run hf.co/mradermacher/gpt-neox-20b-i1-GGUF:Q5_K_MDownload hf download mradermacher/gpt-neox-20b-i1-GGUF --include "gpt-neox-20b.i1-Q5_K_M.gguf" --local-dir gpt-neox-20b-i1-GGUF |
| Q6_K | 15.7 GBfile size | 24.5 GB15.7 + 8.3 + 0.50 | Commandsllama.cpp llama-server -hf mradermacher/gpt-neox-20b-i1-GGUF:Q6_KOllama ollama run hf.co/mradermacher/gpt-neox-20b-i1-GGUF:Q6_KDownload hf download mradermacher/gpt-neox-20b-i1-GGUF --include "gpt-neox-20b.i1-Q6_K.gguf" --local-dir gpt-neox-20b-i1-GGUF |
MLX for Apple silicon
| Format | Size | Memory needed | Repo |
|---|---|---|---|
| MLX 4bitnhe-ai/gpt-neox-20b-mlx-4Bit | 10.8 GBfile size | 19.5 GB10.8 + 8.3 + 0.50 | Download page |
Original weights size from the safetensors header, for vLLM, SGLang, Transformers
| Format | Size | Memory needed | Repo |
|---|---|---|---|
| F16original weights | 38.5 GBfile size | 47.2 GB38.5 + 8.3 + 0.50 |
hf download EleutherAI/gpt-neox-20b --local-dir gpt-neox-20bOther conversions (1; sizes not read)
- OpenVINO/gpt-neox-20b-int8-ov Other, int8, 44 downloads
The Hub tags 9 quantised repos as conversions of this model; we read the most downloaded and show the maker's own first.
Run it by API
No provider on Hugging Face Inference Providers serves this model (as of 19 Nov 2026, 13:58 UTC). Other hosts may; we only list what the router reports.
Not built yet: one key for every open model, with BlockRun, and our own tested-quality score per provider. Until then, precision and measured speed are the best guide to a provider here.
Planned for this page Not built yet
- Try it in private chat
- A link to Solana AI's confidential chat for the models it serves. This model is not listed as served there.
- Fine-tune it
- Fine-tuning a few base models from this page and publishing the result here.
- Creators and earnings
- Pages for people who publish fine-tunes and adapters of this model, and what they earn.
- Fund and bounties
- Bounties for new open models, paid from a public fund. How the fund will work.
- Contribute compute
- How to give compute to open model work, with Akash.
Model data from the Hugging Face Hub 19 Nov 2026, 13:43 UTC. The repo. Found a mistake on this page? Tell us.