Second pre-alpha from Tyness Train, an agent-guided fine-tuning harness for small LLMs on low-VRAM GPUs. This model is RAG-style grounded: every prompt carries excerpts of the relevant norm, and the model is trained to answer only from that context. Three iterations established that SFT alone teaches format, not facts on 1.2B models: legalize-es, with teacher-generated ChatML entries from es-legal-priv. 10 BOE articles, 1,148 train + 20 held-out eval items: - Each entry cites specific precepts (articuloscitados); those sections are extracted from the norm markdown and embedded as a CONTEXTO block in the user turn (up to 9,000 chars). - The system prompt enforces grounded behavior: answer…
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.
11,122Models
3,164Datasets
285Papers
6,792Publishers
11,852Sourced relationships
Updated 2026-10-18 · How the library is built
11,122 models, sorted by most downloaded.
Third pre-alpha from Tyness Train, an agent-guided fine-tuning harness for small LLMs on low-VRAM GPUs. This model is RAG-style grounded (answers only from provided norm excerpts) and adds two confirmed recipe upgrades: a structured METADATOS block in the system prompt (official identifiers, dates, rank, department, article-title index) and adversarial correction pairs (user asserts a false premise → the model corrects it with the exact citation). Evaluated on fresh, uncontaminated teacher-generated sets: 200 questions per context style (never part of any training split), each grounded in a specific precept, judged by an LLM referee (glm5.3-flash): correct / partial / incorrect + 0–1 score.…
A specialist for the Typed Decisions (workflow, question) on a frozen sentence-transformers/all-MiniLM-L6-v2 encoder, 20 heads in total. Each head takes the rendered state, the question text and its options, and returns a probability distribution over the question's labels. This is not zero-shot. It was fitted on the 1,200-case train split of Typed Decisions and is scored on the 400-case test split. Per the benchmark's own rules it is a specialist, and should not be compared with generalist models that answer unseen question schemas. "This run" is the model in this repository, fitted and scored from the public dataset with adaptive-classifier 0.2.0 on Apple MPS, and scored by the benchmark…
Trackformer 1.2 is a research-only Western Pacific tropical-cyclone forecast model. It evolves a sea-level-pressure (MSLP) field and a moving storm-centred pressure core every six hours through +120 h. A track and central-pressure estimate are extracted from the evolving core, not independently drawn on top of a pressure image. The prior Trackformer 1.1 release remains available and unchanged. Not an operational warning system. Do not use these forecasts for evacuation, aviation, maritime, or other safety-critical decisions. The public 1.2 name identifies one selected development checkpoint; a genuinely untouched storm-level holdout has not yet established generalization. The two models use…
LoRA adapter for google/gemma-4-E4B-it fine-tuned for bidirectional Russian ↔ Tuvan translation. Tuvan is a low-resource Turkic language with roughly 280,000 speakers. On the clean held-out test set this adapter significantly outperforms the specialized translation model Agisight/nllb-rus-tyv-v3-dict16.5k (NLLB-200-distilled-600M fine-tuned on the same data) in both directions and on every metric. Clean test set — 1,369 pairs. The held-out test split used for NLLB v3 (shuffle(seed=42), pairs 2000–4000 of Agisight/tyv-rus-200k), with pairs whose Tuvan or Russian side also appears in training removed. Greedy decoding, same length limit as NLLB (32 + 3 × input length new tokens). Full NLLB…
Fine-tuned Ukrainian word-sense-disambiguation model. - Full MTEB task-level results are in evaluation/mtebresults/.
Fine-tuned Ukrainian word-sense-disambiguation model. - Full MTEB task-level results are in evaluation/mtebresults/.
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). - PEFT 0.18.1
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). - PEFT 0.20.0
Held-out exact IR reconstruction: 0/2. Native structural validity: 2/2. These two authored paraphrase examples share semantic targets with training; they are not independent real-world benchmark tasks. The model generates plausible but semantically incorrect candidates and must not authorize actions. This is an opt-in development artifact. It is not the default symbolic compiler, not a verified formalizer, and not a security repair policy. All qualification, admission and proof-authority flags remain false. No LLM/provider fallback runs. Input embeddings use thenlper/gte-small at commit 17e1f347d17fe144873b1201da91788898c639cd (384 dimensions). The embedding encoder weights are not…
umba-5.6.600 is a state-of-the-art agentic software engineering model fine-tuned on top of Kwaipilot/KAT-Coder-V2.5-Dev (Sparse MoE 35B total / 3B active parameters). Quantized using LocalAI APEX (Adaptive Precision for EXpert Models)
This qwen35 model was trained 2x faster with Unsloth
A structured set of research notes on Vision Language Pretraining, with concrete evaluation references and open questions. Plans and hypotheses are kept separate from completed results. - the scope of the research question and likely confounders - a proposed comparison with matched baselines - concrete evaluation context such as task-appropriate public benchmarks named in the main note - reproducibility checks, failure modes, and open questions - topic-relevant references Start with reading.md for the full note. Sections labeled as plans or hypotheses should not be interpreted as experimental results. If results are added later, they should include dataset versions, commands, seeds…
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.