SAVRN Model Hub
Open-Weight Models
An open-weight model is an AI model whose trained weights are published for anyone to download. The weights are what the model learned in training. With a copy of them you can run the model on hardware you control and train it further on your own data.
Open weights are not the same as open source. Many publishers release the weights without the training data or code, and the license sets what you may do with the model. This library puts each model's full card, architecture, files, license and published evaluations on one page.
12,579Models
3,583Datasets
291Papers
7,578Publishers
12,850Sourced relationships
Updated 2026-10-23 · How the library is built
12,579 models, sorted by most downloaded.
A 3.91bpw quant of DeepSeek-V4.1-Flash-1017 for DwarfStar (ds4), with 2.25bpw routed experts and 8-bit or higher everything else. It uses the same recipe and file layout as antirez's DeepSeek-V4.1-Flash-Q2.gguf, built from the 1017 weights with an imatrix calibrated on 1017. Designed for SSD-streaming on 128GB systems, where decode can reach up to 20tps with expert caching. The language GGUF includes 188.83 GiB of native FP8 Engram tables. ds4 reads the rows it needs directly from disk and never loads the whole table, so keep the file on a fast local SSD. The language model loads in current ds4 with no changes. Memory options are the same as for antirez's Q2: SSD streaming on one 128 GB…
The EXL3 quantization made by Mia-AiLab of deepseek-ai/DeepSeek-V4.1-Flash, repackaged as GGUF without re-quantization. Requires ranma.cpp ranma20261013 or newer (github.com/saturnsky/ranma.cpp).
The engram tables of deepseek-ai/DeepSeek-V4.1-Flash, copied from the original FP8 checkpoint without changing a value. Requires ranma.cpp ranma20261013 or newer (github.com/saturnsky/ranma.cpp).
The current publication is in source/: original Engram safetensors copied server-side from deepseek-ai/DeepSeek-V4.1-Flash, with the upstream license, provenance and a pinned engram-manifest.json TP2 preparation recipe. The DeepSeek41 runtime prepares only its local rank, copying the original FP8 weight and scale bytes without numerical conversion. Full-file SHA256 checks must match all four TP2 packs from the accepted 192.168.1.160 / 192.168.1.134 deployment. These raw safetensors require the source-aware runtime; they are not directly consumable by the Engram pack reader. Older split-part files, if present at the repository root, are legacy artifacts. Use the exact files and revisions…
BertholomusAI recipe: DeepSeek-V4.1-Flash served tensor-parallel over two NVIDIA DGX Spark / GB10 nodes by a deepseekv41 model family we wrote for TensorFold (by ashhart). Clean-room: the model math is re-implemented from DeepSeek's MIT inference code and tech report; no code from other DeepSeek-V4.1 recipes or kits was read or copied (ATTRIBUTION.md). Decode and prefill are what we worked on this time. The same release suite on the served lane for both builds (v0.4 = engine bd0024d at --context 262144, v0.5 = 808eb4a at --context 1048576), same client (tools/dsv41/kitbench.py), same prompts, greedy, DSpark drafting. Median of 3 unless noted. four long streams together give the same replies…
with permanent weight-level abliteration — the safety guardrails have been surgically removed while preserving MMLU capability, vision, reasoning, MTP (DSpark), and multi-turn coherence. Proprietary weight-level abliteration developed by the dealignai research team. No custom model.py, no runtime hooks, no steering vectors — it's a standard checkpoint that loads exactly like the base model. The refusal circuitry is surgically removed while every capability-critical component (routed experts, Engram memory, CSA2 sparse attention, DSpark draft head, vision tower, router gates, norms, embeddings) is preserved byte-identical to the base. Every response 4-tier graded (HARDREF / SOFTRED / HEDGE /…
dealignai/DeepSeek-V4.1-Flash-UNCENSORED-EXL3-2.9bpw (abliteration and EXL3 quantization by dealignai), repackaged as GGUF without re-quantization. Requires ranma.cpp ranma20261013 or newer (github.com/saturnsky/ranma.cpp).
A lossless MXFP4-CSF container of deepseek-ai/DeepSeek-V4.1-Flash at revision dba1be0a40aa45a94ad051997016db3960a90277. CSF stores the routed-expert block-scale planes in a compressed form. Nothing else changes: FP4 weight nibbles, every other tensor, the tensor names, dtypes, shapes and the source metadata are all byte-identical. Decoding is integer arithmetic on bytes, with no requantization or fitting. The original safetensors shards can be restored bit-exactly. - Built with trellis-quant trellisquant.losslessscalecheckpoint (commit 60feca330087, family deepseekv41). - verification.json: every original shard was rebuilt from this container, and both the rebuilt and the stored files…
This model was quantized using oQ (oMLX v0.7.1) mixed-precision quantization with Improved oQe Quantization (jundot/omlx#4385). A mixed-precision MLX checkpoint of deepseek-ai/DeepSeek-V4.1-Flash, a multimodal mixture-of-experts model with 552B backbone parameters, a Causal Encoder-Decoder layout, Compressed Sparse Attention 2 (CSA2), Engram conditional memory and DSpark speculative decoding. It includes the language model, the Engram tables, the vision encoder, the multimodal aligner and the DSpark MTP weights. The effective average is 3.104 bits/weight excluding the Engram tables, or 3.592 bits/weight for the complete checkpoint. oQ2.5e allocates bits by measured layer sensitivity, then…
M3 Ultra 512GB + M4 Max 128GB: ~15 tokens/s @ 1000 tokens ~501 GiB This build repacks the majority of the model's existing quantization, with the remainder quantized to 9 bits. This configuration empirically yielded the highest quality in our testing. We are not the creator, originator, or owner of any model listed. Each model is created and provided by third parties. Models may not always be accurate or contextually appropriate. You are responsible for verifying the information before making important decisions. We are not liable for any damages, losses, or issues arising from its use, including data loss or inaccuracies in AI-generated content.
Merged checkpoint produced by the family-aware Delta-P2S experiment package.
HorneSci's open-weight continued-pretraining release for Qwen3.8-27B. DeltaRoutingAI trains language adapters on Python technical documentation and delivers a 4.61% reduction in held-out text NLL in its paired evaluation. The downloadable checkpoint is a PEFT LoRA adapter in safetensors format (233.6 MB). Load it with Qwen/Qwen3.8-27B, pinned to revision 1d4bf0f2ff6012fd82039f2fa52739d0dd7c60c0. Qwen authors the base model; HorneSci authors the continued-pretraining recipe and adapter weights. The adapter weights are released under Apache 2.0 with upstream attribution and data notices included. Use a CUDA environment with the versions in requirements-inference.txt. The evaluated environment…
On-device LiteRT / TFLite conversion of Depth Anything 3 — Small (ByteDance-Seed, Apache-2.0) for monocular depth, running fully on the mobile GPU via the LiteRT CompiledModel API (ML Drift delegate). No CPU fallback ops — the whole graph is GPU-compatible. DA3 processes images at their native aspect ratio (upperboundresize, longer side → 896, multiple of 14). Forcing a square 896×896 and letterbox-padding drops the match to corr 0.977 (the black padding leaks into the content through global attention). Converting at the native rectangle restores corr 0.9994 and is also faster (fewer tokens). This checkpoint is built for portrait ~9:16. For another aspect, re-convert at that shape (or your…
Model Collections
Hand-picked starting points, each with the reason it exists.
Collection · 4 entries
Embedding models for retrieval
Sentence and document embedding models used to build retrieval systems. Dimension and sequence length matter more than size here, and both come from the publisher.
Collection · 4 entries
Models that fit on one accelerator
Models whose publisher-reported parameter count puts them within reach of a single accelerator at common precisions. Memory needed depends on precision and serving configuration, so treat the parameter count as the starting point, not the answer.
Collection · 6 entries
Open-weight text models worth knowing
Widely used open-weight language models, chosen because each one is a distinct family rather than a variant of the one above it. Selection, not a ranking.
Collection · 3 entries
Speech and audio models
Recognition and synthesis models, grouped so the two directions are easy to compare.
Open-Weight Models Explained
What is an open-weight model?
An AI model whose trained weights are published for anyone to download, so it can be run, tested and fine-tuned on hardware the user controls.
Is an open-weight model the same as open source?
Not always. Open weights means the trained model can be downloaded. Open source usually also means the training code and data are available and the license allows broad reuse. Many open-weight models release the weights only.
Can I use an open-weight model commercially?
It depends on the license. Apache 2.0 and MIT allow commercial use. Other licenses limit it, for example to non-commercial use or below a set number of users. Every model page here shows its license.
How much memory does an open-weight model need?
About two bytes per parameter at 16-bit precision, so a 7-billion-parameter model needs roughly 14 GB for its weights, plus memory for the context it processes. Each model page lists its parameter count and the size of its files.
Related SAVRN Research
The hub sits beside SAVRN's market data and infrastructure research: what models cost to run, and what it takes to run them.
SAVRN Index
What open models cost to run
The same open-weight model priced by every host that serves it, per million tokens.
Research Hub
Data center trackers and maps
Moratoriums, permits, power, water and capital behind the facilities that run these models.
Method
How the Model Hub is built
Sources, evidence labels, refresh behaviour, and the limits of every comparison here.