IdeaLens detects who came up with the ideas in a document, rather than who wrote its words. It reads a role-labelled outline of the document (an ordered list of items, each giving one idea and the discourse role it plays, such as Central Development or Open Question) and returns P(human), the probability that the ideas are human. IdeaLens is nvidia/NVIDIA-Nemotron-3.5-Lightning-30B-A3B-BF16 fine-tuned with LoRA (rank 64) on the outlines of 1M English web documents (WildOutlines). The training outlines were paraphrased to remove the documents' wording, so the model has to fit its labels through the ideas. Try it in your browser: the IdeaLens & ProseLens demo scores your own text with both…
ProseLens is an open-weight model for text generation from Rishanth Rajendhran, released under other. It has 32.9B parameters and a 262,144-token context. At 16-bit it needs about 79 GB of GPU memory, which fits on 1x MI300X from $1.85 an hour, at the lowest prices in the SAVRN Index. It draws 11 downloads a month.
ProseLens detects who wrote the words of a document. It is IdeaLens's counterpart in the paper: the same backbone, training documents and labels, but it reads the raw document text instead of an outline, so it learns word-level provenance.
Parameters32.9B
Context262,144
Weights68.9 GB
Licenseother
AccessOpen weights
Monthly Downloads11
Runs On
What it takes to serve ProseLens (32.9B 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 | 65.8 GB | 79.0 GB | 1x MI300X (192 GB) Vultr |
$1.85 | 1x H100 $1.99 · 1x MI325X $2.00 |
| 8-bit | 32.9 GB | 39.5 GB | 1x MI300X (192 GB) Vultr |
$1.85 | 1x H100 $1.99 · 1x MI325X $2.00 |
| 4-bit | 16.5 GB | 19.7 GB | 1x MI300X (192 GB) Vultr |
$1.85 | 1x H100 $1.99 · 1x 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, read Oct 6, 2026.
ProseLens on every accelerator the SAVRN Index prices, at every precision
Model Card
ProseLens detects who wrote the words of a document. It is IdeaLens's counterpart in the paper: the same backbone, training documents and labels, but it reads the raw document text instead of an outline, so it learns word-level provenance. It returns P(human), the probability that the document was written by a person. ProseLens is nvidia/NVIDIA-Nemotron-3.5-Lightning-30B-A3B-BF16 fine-tuned with LoRA (rank 64) on 1M English web documents (WildOutlines). Try it in your browser: the IdeaLens & ProseLens demo scores your own text with both detectors, no installation or keys needed. From the paper: ProseLens is accurate when a document's ideas and words come from the same source (99.1%), but…
Excerpt from the card by Rishanth Rajendhran, licensed other.
Configuration
- Architecture
- NemotronHForCausalLM
- Context length (tokens)
- 262,144
- Layers
- 52
- Hidden size
- 2,688
- Feed-forward size
- 1,856
- Attention heads
- 32
- Key/value heads
- 2
- Head dimension
- 128
- Vocabulary size
- 131,072
- Routed experts
- 128
- Experts active per token
- 6
- RoPE base
- 10,000
- Model type
- nemotron_h
Identity and Version
- Repository
- rishanthrajendhran/ProseLens
- Publisher
- Rishanth Rajendhran
- Task
- Text generation
- Modality
- Text
- Library
- transformers
- Parameters
- 32.9B parameters
- Languages
- en
- Revision
- 2d66972d105051c1bc0115c80401b82d8819f1e4
- First published
- 2026-08-31
- Last updated
- 2026-10-06
Files and Weights
27 files, 68.9 GB in total. The weights are 15 files totalling 68.9 GB in safetensors.
Weights15 files · 68.9 GB
Configuration6 files · 623.8 KB
Tokenizer2 files · 17.3 MB
Documentation2 files · 17.0 KB
Other1 file · 9.9 KB
Repository1 file · 1.6 KB
Every file
| File | Type | Size | SHA-256 |
|---|---|---|---|
| adapter/adapter_model.safetensors | Weights | 3.1 GB | 6aa1965bc0c7 |
| model-00001-of-00014.safetensors | Weights | 5.0 GB | cd128bfd9342 |
| model-00002-of-00014.safetensors | Weights | 5.0 GB | 85969f9b02e6 |
| model-00003-of-00014.safetensors | Weights | 5.0 GB | 3c1cb332c55e |
| model-00004-of-00014.safetensors | Weights | 5.0 GB | 63098215bc45 |
| model-00005-of-00014.safetensors | Weights | 5.0 GB | db7d1f3278be |
| model-00006-of-00014.safetensors | Weights | 5.0 GB | e567c95f1fc5 |
| model-00007-of-00014.safetensors | Weights | 5.0 GB | a9382a9eb90d |
| model-00008-of-00014.safetensors | Weights | 5.0 GB | 0c26027509bb |
| model-00009-of-00014.safetensors | Weights | 5.0 GB | c3cc352129ba |
| model-00010-of-00014.safetensors | Weights | 5.0 GB | 5723b717fbbe |
| model-00011-of-00014.safetensors | Weights | 5.0 GB | 79067ee3d842 |
| model-00012-of-00014.safetensors | Weights | 5.0 GB | e68b02cf6cce |
| model-00013-of-00014.safetensors | Weights | 3.2 GB | a0ec01201aa1 |
| model-00014-of-00014.safetensors | Weights | 2.7 GB | 64577b275ca4 |
| adapter/adapter_config.json | Configuration | 982 B | — |
| config.json | Configuration | 2.4 KB | — |
| load_adapter.py | Configuration | 5.1 KB | — |
| model.safetensors.index.json | Configuration | 612.2 KB | — |
| special_tokens_map.json | Configuration | 563 B | — |
| thresholds.json | Configuration | 2.6 KB | — |
| LICENSE | Documentation | 2.7 KB | — |
| README.md | Documentation | 14.3 KB | — |
| chat_template.jinja | Other | 9.9 KB | — |
| .gitattributes | Repository | 1.6 KB | — |
| tokenizer.json | Tokenizer | 17.1 MB | 623c34567aeb |
| tokenizer_config.json | Tokenizer | 177.2 KB | — |
License and Download
- License
- other
- Access
- Open weights, no gate
- Download size
- 68.9 GB
Download from Rishanth Rajendhran
Released by Rishanth Rajendhran through its official repository on Hugging Face.
Built From
- Derived from nvidia/NVIDIA-Nemotron-3.5-Lightning-30B-A3B-BF16
- Described by arXiv:2610.06778
- Trained on (disclosed) rishanthrajendhran/WildOutlines
Memory Requirements
| Precision | Weights in memory |
|---|---|
| As published | 68.9 GB |
| 16-bit | 65.8 GB |
| 8-bit | 32.9 GB |
| 4-bit | 16.5 GB |
Weights only, from the published parameter count; the key-value cache and runtime add to this.
Questions About ProseLens
How much GPU memory does ProseLens need?
About 79 GB at 16-bit and 19.7 GB at 4-bit: the weights (32.9B parameters) plus a working margin. A long context needs more.
What is the cheapest GPU to run ProseLens 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.
What license is ProseLens released under?
other, as its publisher declares it. Read the license text before commercial use.
What is ProseLens's context length?
262,144 tokens, from the maximum position embeddings in its published configuration.
Similar Models
IdeaLens-NoParaphrase is an ablation of IdeaLens: the same backbone, documents and labels, trained on the outlines as extracted, without the paraphrasing step that removes the documents' wording. It reads a role-labelled outline and returns P(human), the probability that the document's ideas are human. It is nvidia/NVIDIA-Nemotron-3.5-Lightning-30B-A3B-BF16 fine-tuned with LoRA (rank 64) on the outlines of 1M English web documents (WildOutlines, outline field). Try it in your browser: the IdeaLens & ProseLens demo scores your own text with both detectors, no installation or keys needed. This model is not reported in the paper. It is released for comparison with IdeaLens. Scoring a document…
Qwen2.5 is the latest series of Qwen large language models. For Qwen2.5, we release a number of base language models and instruction-tuned language models ranging from 0.5 to 72 billion parameters. Qwen2.5 brings the following improvements upon Qwen2: - Significantly more knowledge and has greatly improved capabilities in coding and mathematics, thanks to our specialized expert models in these domains. - Significant improvements in instruction following, generating long texts (over 8K tokens), understanding structured data (e.g, tables), and generating structured outputs especially JSON. More resilient to the diversity of system prompts, enhancing role-play implementation and…
Qwen2.5 is the latest series of Qwen large language models. For Qwen2.5, we release a number of base language models and instruction-tuned language models ranging from 0.5 to 72 billion parameters. Qwen2.5 brings the following improvements upon Qwen2: - Significantly more knowledge and has greatly improved capabilities in coding and mathematics, thanks to our specialized expert models in these domains. - Significant improvements in instruction following, generating long texts (over 8K tokens), understanding structured data (e.g, tables), and generating structured outputs especially JSON. More resilient to the diversity of system prompts, enhancing role-play implementation and…
Qwen2.5-Coder is the latest series of Code-Specific Qwen large language models (formerly known as CodeQwen). As of now, Qwen2.5-Coder has covered six mainstream model sizes, 0.5, 1.5, 3, 7, 14, 32 billion parameters, to meet the needs of different developers. Qwen2.5-Coder brings the following improvements upon CodeQwen1.5: - Significantly improvements in code generation, code reasoning and code fixing. Base on the strong Qwen2.5, we scale up the training tokens into 5.5 trillion including source code, text-code grounding, Synthetic data, etc. Qwen2.5-Coder-32B has become the current state-of-the-art open-source codeLLM, with its coding abilities matching those of GPT-4o. - A more…
Qwen2.5-Coder is the latest series of Code-Specific Qwen large language models (formerly known as CodeQwen). As of now, Qwen2.5-Coder has covered six mainstream model sizes, 0.5, 1.5, 3, 7, 14, 32 billion parameters, to meet the needs of different developers. Qwen2.5-Coder brings the following improvements upon CodeQwen1.5: - Significantly improvements in code generation, code reasoning and code fixing. Base on the strong Qwen2.5, we scale up the training tokens into 5.5 trillion including source code, text-code grounding, Synthetic data, etc. Qwen2.5-Coder-32B has become the current state-of-the-art open-source codeLLM, with its coding abilities matching those of GPT-4o. - A more…