Public adapter release for dLLM PRM Gap. a C3 control checkpoint with the stated attention/readout protocol and seed. - adapter.safetensors — compact adapter weights - config.json — public base-model id, architecture, and provenance
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.
11,734Models
3,323Datasets
288Papers
7,150Publishers
12,282Sourced relationships
Updated 2026-10-20 · How the library is built
11,734 models, sorted by most downloaded.
Public adapter release for dLLM PRM Gap. causal attention with a last-token readout for the submitted readout diagnostic - adapter.safetensors — compact adapter weights - config.json — public base-model id, architecture, and provenance Role: last-token scorer used for the last-token row in the readout diagnostic.
Public adapter release for dLLM PRM Gap. causal PRM for the cross-backbone LLaDA diagnostic - adapter.safetensors — compact adapter weights - config.json — public base-model id, architecture, and provenance Role: cross-backbone diagnostic; it is not a full LLaDA ORM replication.
Public adapter release for dLLM PRM Gap. causal attention with mask-aware mean pooling for the submitted readout diagnostic - adapter.safetensors — compact adapter weights - config.json — public base-model id, architecture, and provenance Role: causal mean-pool scorer used for the mean-pool row in the readout diagnostic.
Public adapter release for dLLM PRM Gap. bidirectional final-state scoring for the matched-compute ORM Rerank baseline - adapter.safetensors — compact adapter weights - config.json — public base-model id, architecture, and provenance
Pipeline Class: CraftlyImagePipeline; Transformer Diffusion Backbone: CraftlyImageTransformer2DModel (28 Layers, 48 Attention Heads, 64-channel Latent Input); Text Conditioning Backbone: CraftlyVisionModel (36 Layers, 2560 Hidden Dim); Latent Autoencoder: 16-channel Spatial VAE; Denoising Trajectory: Flow-Matching Euler Discrete Scheduler (Optimized 8-step Distillation).
Model · Text generation
dan-100mb-after-wc-loglinear-newlex-before-ckpt500-packed-bfdiso_seed10
This model is a fine-tuned version of francesca9805/ppt-wc-loglinear-newlex-dan-before-100mb-packed-bfdisoseed10. It has been trained using TRL. This model was trained with SFT.
Model · Text generation
dan-100mb-after-wc-loglinear-newlex-before-ckpt500-packed-bfdiso_seed3407
This model is a fine-tuned version of francesca9805/ppt-wc-loglinear-newlex-dan-before-100mb-packed-bfdisoseed3407. It has been trained using TRL. This model was trained with SFT.
Daniel is an autonomous AI software engineer built by Cognitive AI. It reads your codebase, writes code, runs tests, and opens pull requests — without you doing anything manually. Daniel was tested head-to-head against Devin on 6 real bugs from popular open-source projects. Both agents received identical prompts. Each PR was graded by the project's own hidden test suite. Daniel is 70% cheaper than Devin at the same pass rate. Daniel uses a 4-turn agentic loop: 1. Analyze — reads the issue, the failing tests, and the relevant code 2. Implement — makes the minimal code change to fix the bug 3. Test — runs the project's existing test suite to verify the fix 4. PR — opens a pull request with a…
Fresh LoRA rank 16/alpha 32 on the same 2048-question split used by the STRIDE comparisons. Explicit nonthinking (enablethinking=False), seed42, eight responses per question, context 8192, and four learner plus four sampler GPUs. This uses DAPO asymmetric clipping, dynamic filtering/refill of mixed-correctness groups, global token-normalized policy loss, and a soft overlength reward. The exact released-code provenance and implemented equations are recorded below. Unlike the original paper's no-KL recipe, this controlled comparison uses the shared KL coefficient 0.01 against the frozen initial base. This is not an exact reproduction of the paper's scale or training recipe and makes no…
Fresh LoRA rank 16/alpha 32 on the same 2048-question split used by the STRIDE comparisons. Explicit nonthinking (enablethinking=False), seed42, eight responses per question, context 8192, and four learner plus four sampler GPUs. This uses DAPO asymmetric clipping, dynamic filtering/refill of mixed-correctness groups, global token-normalized policy loss, and a soft overlength reward. The exact released-code provenance and implemented equations are recorded below. Unlike the original paper's no-KL recipe, this controlled comparison uses the shared KL coefficient 0.01 against the frozen initial base. This is not an exact reproduction of the paper's scale or training recipe and makes no…
A structured set of research notes on Data Efficient Learning, 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 analysis.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, hardware…
Reading notes and an experiment sketch for Data Efficient Learning. The repository emphasizes what still needs to be tested instead of manufacturing scores or release claims. - 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 papernotes.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, hardware, and…
Released checkpoints and evaluation keypoints for RoboTok, a model that retrieves web video clips by 3D hand-motion similarity. Similarity is defined by DTW over torso-relative 3D hand keypoints; the encoder is trained to reproduce that DTW ranking in a fast embedding space. Training and evaluation code is in the accompanying source release. FAIR Noncommercial Research License v1 (see LICENSE-Action100M). Noncommercial research only. The released checkpoints (models/.pt) and evaluation keypoints (evaldata/torsorelativeclipkeypoints.pt) are derivative works of Action100M (Meta FAIR) clips and are governed by that license. It covers trained model weights as "Research Materials", and restricts…
Original diagrams for this release. Editable SVG versions: architecture · inference and distillation. Weights and completed test results are included in this release bundle. The detector compares encoder blocks 1, 2 and 3 with the final posterior-mean representation of a frozen FLUX.2 VAE. Each stream is projected to an 8 × 8 token grid. The three early-to-final differences are fused with all four feature streams before transformer classification. Inputs are native-resolution 128 × 128 RGB crops with a central 96 × 96 active region. The VAE decoder and generative FLUX transformer are not used. The training run excludes CelebA-HQ from training, validation and patch caches; it is used only…
Original diagrams for this release. Editable SVG versions: architecture · inference and distillation. Weights and completed test results are included in this release bundle. The detector compares encoder blocks 1, 2 and 3 with the final posterior-mean representation of a frozen FLUX.2 VAE. Each stream is projected to an 8 × 8 token grid. The three early-to-final differences are fused with all four feature streams before transformer classification. Inputs are native-resolution 128 × 128 RGB crops with a central 96 × 96 active region. The VAE decoder and generative FLUX transformer are not used. The training run excludes CelebA-HQ from training, validation and patch caches; it is used only…
Original diagrams for this release. Editable SVG versions: architecture · inference and distillation. Weights and completed test results are included in this release bundle. The detector compares encoder blocks 1, 2 and 3 with the final posterior-mean representation of a frozen FLUX.2 VAE. Each stream is projected to an 8 × 8 token grid. The three early-to-final differences are fused with all four feature streams before transformer classification. Inputs are native-resolution 128 × 128 RGB crops with a central 96 × 96 active region. The VAE decoder and generative FLUX transformer are not used. The training run excludes CelebA-HQ from training, validation and patch caches; it is used only…
Original diagrams for this release. Editable SVG versions: architecture · inference and distillation. Weights and completed test results are included in this release bundle. The detector compares encoder blocks 1, 2 and 3 with the final posterior-mean representation of a frozen FLUX.2 VAE. Each stream is projected to an 8 × 8 token grid. The three early-to-final differences are fused with all four feature streams before transformer classification. Inputs are native-resolution 128 × 128 RGB crops with a central 96 × 96 active region. The VAE decoder and generative FLUX transformer are not used. The training run excludes CelebA-HQ from training, validation and patch caches; it is used only…
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.