# pythia-6.9b by EleutherAI: Open-Weight Model Source: https://savrn.com/models/pythia-6-9b Markdown alternate of the page above; the site index is https://savrn.com/llms.txt --- ## Runs On What it takes to serve pythia-6.9b (7B 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 | 14.0 GB | 16.8 GB | 1x [MI300X](https://savrn.com/ai-index/pricing/gpus/mi300x) (192 GB) Vultr | $1.85 | [1x H100](https://savrn.com/ai-index/pricing/gpus/h100) $1.99 · [1x MI325X](https://savrn.com/ai-index/pricing/gpus/mi325x) $2.00 | | 8-bit | 7.0 GB | 8.4 GB | 1x [MI300X](https://savrn.com/ai-index/pricing/gpus/mi300x) (192 GB) Vultr | $1.85 | [1x H100](https://savrn.com/ai-index/pricing/gpus/h100) $1.99 · [1x MI325X](https://savrn.com/ai-index/pricing/gpus/mi325x) $2.00 | | 4-bit | 3.5 GB | 4.2 GB | 1x [MI300X](https://savrn.com/ai-index/pricing/gpus/mi300x) (192 GB) Vultr | $1.85 | [1x H100](https://savrn.com/ai-index/pricing/gpus/h100) $1.99 · [1x MI325X](https://savrn.com/ai-index/pricing/gpus/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](https://savrn.com/ai-index/pricing/gpus), read Oct 23, 2026. [pythia-6.9b on every accelerator the SAVRN Index prices, at every precision](https://savrn.com/models/pythia-6-9b/gpus) ## SAVRN's Notes on pythia-6.9b 16.8 GB at 16-bit, 8.4 GB at 8-bit, 4.2 GB at 4-bit, 7B parameters — and a 2,048-token context that is the real constraint. Pythia predates the long-context era, and no quantization changes that. The cheapest Index rental is one MI300X at $1.85 an hour, comfortably sufficient, though the model is small enough for high-end workstation cards you may already own. What you are buying is research infrastructure: the Pythia suite trains 8 sizes on identical data in identical order, with a deduplicated twin set and 154 intermediate checkpoints published as branches, which is why the repository holds 27.7 GB across four files (safetensors and pytorch copies — take one). Apache 2.0 permits commercial use, but this is a base model with no chat tuning and no evaluations on this page. For production assistants pick a modern instruct model; for interpretability or scaling studies, nothing at this price substitutes. SAVRN Research ## Model Card By EleutherAI, published under apache-2.0, revision c0e3eee36dc4. The Pythia Scaling Suite is a collection of models developed to facilitate interpretability research [(see paper)](https://savrn.com/papers/pythia-a-suite-for-analyzing-large-language-models-across-training-and-scaling). It contains two sets of eight models of sizes 70M, 160M, 410M, 1B, 1.4B, 2.8B, 6.9B, and 12B. For each size, there are two models: one trained on the Pile, and one trained on the Pile after the dataset has been globally deduplicated. All 8 model sizes are trained on the exact same data, in the exact same order. We also provide 154 intermediate checkpoints per model, hosted on Hugging Face as branches. The Pythia model suite was deliberately designed to promote scientific research on large language models, especially interpretability research. Despite not centering downstream performance as a design goal, we find the models match or exceed the performance of similar and same-sized models, such as those in the OPT and GPT-Neo suites. [Read the full model card (1,371 words)](https://savrn.com/models/pythia-6-9b/card) ## Configuration Architecture GPTNeoXForCausalLM Context length (tokens) 2,048 Layers 32 Hidden size 4,096 Feed-forward size 16,384 Attention heads 32 Vocabulary size 50,432 Stored precision float16 Model type gpt_neox ## Identity and Version Repository EleutherAI/pythia-6.9b Publisher EleutherAI Task Text generation Modality Text Library transformers Parameters 7B parameters Languages en Revision c0e3eee36dc47af0c49f361c74cfe459c09f7f23 First published 2023-02-14 Last updated 2025-03-10 ## Files and Weights 12 files, 27.7 GB in total. The weights are 4 files totalling 27.7 GB in bin, safetensors. Weights4 files · 27.7 GB Configuration4 files · 86.7 KB Tokenizer2 files · 2.1 MB Documentation1 file · 13.6 KB Repository1 file · 1.5 KB Every file | File | Type | Size | SHA-256 | | --- | --- | --- | --- | | model-00001-of-00002.safetensors | Weights | 9.9 GB | 34782c3cf3cd | | model-00002-of-00002.safetensors | Weights | 3.9 GB | f73019e3a2ce | | pytorch_model-00001-of-00002.bin | Weights | 9.9 GB | 753cc06e909d | | pytorch_model-00002-of-00002.bin | Weights | 3.9 GB | 1fde0bc4136f | | config.json | Configuration | 571 B | — | | model.safetensors.index.json | Configuration | 44.0 KB | — | | pytorch_model.bin.index.json | Configuration | 42.0 KB | — | | special_tokens_map.json | Configuration | 99 B | — | | README.md | Documentation | 13.6 KB | — | | .gitattributes | Repository | 1.5 KB | — | | tokenizer.json | Tokenizer | 2.1 MB | — | | tokenizer_config.json | Tokenizer | 396 B | — | ## License and Download License apache-2.0 Access Open weights, no gate Download size 27.7 GB [Download from EleutherAI](https://huggingface.co/EleutherAI/pythia-6.9b) Released by EleutherAI through its official repository on Hugging Face. [Read the license](https://www.apache.org/licenses/LICENSE-2.0). ## Built From - Described by [arXiv:2101.00027](https://savrn.com/papers/the-pile-an-800gb-dataset-of-diverse-text-for-language-modeling) - Described by [arXiv:2201.07311](https://savrn.com/papers/datasheet-for-the-pile) - Described by [arXiv:2304.01373](https://savrn.com/papers/pythia-a-suite-for-analyzing-large-language-models-across-training-and-scaling) - Trained on (disclosed) EleutherAI/pile ## Memory Requirements | Precision | Weights in memory | | --- | --- | | As published | 27.7 GB | | 16-bit | 14.0 GB | | 8-bit | 7.0 GB | | 4-bit | 3.5 GB | Weights only, from the published parameter count; the key-value cache and runtime add to this. ## Questions About pythia-6.9b ### How much GPU memory does pythia-6.9b need? About 16.8 GB at 16-bit and 4.2 GB at 4-bit: the weights (7B parameters) plus a working margin. A long context needs more. ### What is the cheapest GPU to run pythia-6.9b 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 pythia-6.9b commercially? Yes. pythia-6.9b is released under Apache License 2.0. The Apache License 2.0 is a permissive open-source license. It permits commercial use, modification and redistribution. It requires keeping the license and copyright notices and any NOTICE file, stating significant changes, and it includes an express patent grant from contributors. ### What is pythia-6.9b's context length? 2,048 tokens, from the maximum position embeddings in its published configuration. ## Similar Models Model · Text generation ### [granite-4.0-h-tiny-w8a8-llmcompressor](https://savrn.com/models/granite-4-0-h-tiny-w8a8-llmcompressor) [AMD](https://savrn.com/model-publishers/amd) ZenDNN v6.1.0 - ZenTorch v2.13.0.0 - PyTorch v2.13.0.0 - LLM Compressor v0.13.0 - vLLM v0.29.0 This is a quantized version of granite-4.0-h-tiny created by AMD using LLM Compressor (compressed-tensors) for ZenDNN-optimized CPU inference. The model was quantized from granite-4.0-h-tiny using LLM Compressor via the Round-to-Nearest (RTN) algorithm. This reduces the model weights from 12.9 GiB to 6.6 GiB on disk (~49% reduction). granite-4.0-h-tiny is a hybrid Mamba-MoE model: of its 40 layers, 4 are full-attention blocks and the other 36 are Mamba (linear-attention) blocks, and every layer carries a 64-expert MoE block alongside a shared MLP. The recipe only needs two ignore entries. lmhead… Open weights apache-2.0 6.9B parameters 131,072 tokens transformers [View model](https://savrn.com/models/granite-4-0-h-tiny-w8a8-llmcompressor) Model · Text generation ### [deepseek-coder-7b-instruct-v1.5](https://savrn.com/models/deepseek-coder-7b-instruct-v1-5) [DeepSeek](https://savrn.com/model-publishers/deepseek-ai) Deepseek-Coder-7B-Instruct-v1.5 is continue pre-trained from Deepseek-LLM 7B on 2T tokens by employing a window size of 4K and next token prediction objective, and then fine-tuned on 2B tokens of instruction data. Here give some examples of how to use our model. This code repository is licensed under the MIT License. The use of DeepSeek Coder models is subject to the Model License. DeepSeek Coder supports commercial use. See the LICENSE-MODEL for more details. If you have any questions, please raise an issue or contact us at service@deepseek.com. Open weights other 6.9B parameters 4,096 tokens transformers [View model](https://savrn.com/models/deepseek-coder-7b-instruct-v1-5) Model · Text generation ### [Mistral-7B-Instruct-v0.2](https://savrn.com/models/mistral-7b-instruct-v0-2) [Mistral AI_](https://savrn.com/model-publishers/mistralai) The Mistral-7B-Instruct-v0.2 Large Language Model (LLM) is an instruct fine-tuned version of the Mistral-7B-v0.2. Mistral-7B-v0.2 has the following changes compared to Mistral-7B-v0.1 - 32k context window (vs 8k context in v0.1) - Rope-theta = 1e6 For full details of this model please read our paper and release blog post. In order to leverage instruction fine-tuning, your prompt should be surrounded by [INST] and [/INST] tokens. The very first instruction should begin with a begin of sentence id. The next instructions should not. The assistant generation will be ended by the end-of-sentence token id. This format is available as a chat template via the applychattemplate() method: - If you… Open weights apache-2.0 7.2B parameters 32,768 tokens transformers [View model](https://savrn.com/models/mistral-7b-instruct-v0-2) Model · Text generation ### [svd-safety-mis7_swift_remove30_swapgapiter_b010](https://savrn.com/models/svd-safety-mis7-swift-remove30-swapgapiter-b010) [Park](https://savrn.com/model-publishers/jeesup) A Mistral-7B-Instruct-v0.2 checkpoint compressed with SVD-LLM to 70.0% of dense parameters, then edited by 10 of 10 rounds of iterative parameter-neutral swap selected by the gapiter rule (up to 0.1% of dense parameters per round; the full run's budget is 1.0%). This is a research artifact from a study of how SVD compression damages safety behaviour and which component-selection rule best repairs it. It is one cell of a grid over selection rules and budgets; it is not a general-purpose chat model. This checkpoint exists to measure safety/utility trade-offs under compression. Several arms in the grid are deliberately safety-degraded relative to Mistral-7B-Instruct-v0.2: compression alone… Open weights apache-2.0 7.2B parameters 32,768 tokens transformers [View model](https://savrn.com/models/svd-safety-mis7-swift-remove30-swapgapiter-b010) Model · Text generation ### [mistral-7b-qlora-alpaca-sample-0.5k](https://savrn.com/models/mistral-7b-qlora-alpaca-sample-0-5k) [Seongmin Kim](https://savrn.com/model-publishers/mrseongminkim) 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). Open weights 7.2B parameters 32,768 tokens transformers [View model](https://savrn.com/models/mistral-7b-qlora-alpaca-sample-0-5k) Model · Text generation ### [yejin-korean-8b-v3-cpt6000-distill-qwen36-lora-lr2e4-nothink-v2](https://savrn.com/models/yejin-korean-8b-v3-cpt6000-distill-qwen36-lora-lr2e4-nothink-v2) [Hajunho](https://savrn.com/model-publishers/cooler8) Korean instruction-following language model. Built from Yejin 8B v3 (pretrained from scratch by Yejin: cooler8/yejin-korean-8b-v3-base): continued pretraining for 6,000 steps on public Korean data, Yejin chat vector from our own SFT model, then LoRA (rank 64, lr 2e-4). The LoRA data mixes our phase-1 format data (45,013 samples) with 39,290 short Korean solutions written by Qwen/Qwen3.6-27B (Apache-2.0) for questions from our own puzzle generator, KMMLU train, AI Hub school exams and AI Hub reading comprehension; only solutions that reached the known answer were kept (sequence-level distillation; no teacher weights or logits). No benchmark test items were used, and the teacher solutions… Open weights apache-2.0 7.2B parameters 4,096 tokens [View model](https://savrn.com/models/yejin-korean-8b-v3-cpt6000-distill-qwen36-lora-lr2e4-nothink-v2) ## EleutherAI [All models and datasets](https://savrn.com/model-publishers/eleutherai) ## Versions - [c0e3eee36dc4](https://savrn.com/models/pythia-6-9b/versions/c0e3eee36dc4) · current 2026-09-18 ## Explore More - [All text generation models](https://savrn.com/models/tasks/text-generation) - [All models under apache-2.0](https://savrn.com/models/licenses/apache-2-0) - [Model comparisons](https://savrn.com/models/comparisons) - [The model directory](https://savrn.com/models) - [Open model prices by host](https://savrn.com/ai-index/pricing/open-models) ## Source - Repository metadata, read 2026-10-19. - [EleutherAI website](https://www.eleuther.ai/) - [Hugging Face record](https://huggingface.co/EleutherAI/pythia-6.9b) - [How the hub is built](https://savrn.com/model-hub/methodology)