SAVRN Model Hub · Comparisons
deberta-v2-large-japanese-char-wwm vs roberta-large
Deberta-v2-large-japanese-char-wwm has 330M parameters and roberta-large has 355M parameters; deberta-v2-large-japanese-char-wwm is released under Creative Commons Attribution-ShareAlike 4.0 and roberta-large under MIT License; at 16-bit, deberta-v2-large-japanese-char-wwm needs about 0.8 GB (1x MI300X from $1.85 an hour) and roberta-large about 0.9 GB (1x MI300X from $1.85 an hour).
| Field | deberta-v2-large-japanese-char-wwm ku-nlp/deberta-v2-large-japanese-char-wwm | roberta-large FacebookAI/roberta-large |
|---|---|---|
| Publisher | Language Media Processing Lab at Kyoto University | Facebook AI community |
| Task | Fill mask | Fill mask |
| Modality | Text | Text |
| Parameters, as reported | 330M parameters | 355M parameters |
| Architecture | DebertaV2ForMaskedLM | RobertaForMaskedLM |
| Library | transformers | transformers |
| Context length | 512 tokens | 514 tokens |
| Repository size | 2.6 GB | 7.5 GB |
| Artifact formats | safetensors, pytorch | safetensors, onnx, pytorch, jax, tf |
| License | cc-by-sa-4.0 | mit |
| Access | Open weights, no gate | Open weights, no gate |
| Memory at 16-bit (weights and margin) | 0.8 GB | 0.9 GB |
| Cheapest GPUs at 16-bit, per hour | 1x MI300X, $1.85 | 1x MI300X, $1.85 |
| Memory at 4-bit (weights and margin) | 0.2 GB | 0.2 GB |
| Cheapest GPUs at 4-bit, per hour | 1x MI300X, $1.85 | 1x MI300X, $1.85 |
| Revision viewed | 547b0e8b044f | 722cf37b1afa |
| Downloads reported by the hub | 364.3k | 6.5M |
| Last observed | 2026-09-18 | 2026-09-18 |
An evaluation row appears only where at least two of these models report the same benchmark with the same stated configuration, metric, unit and setup. Different evaluators stay named in each cell. Values are shown as reported: no unit conversion, no ranking.
SAVRN's Notes on roberta-large
Loaded at 16-bit, roberta-large needs 0.9 GB of memory for 0.7 GB of weights, and the cheapest setup the SAVRN Index prices for it is one MI300X with 192 GB at $1.85 per hour on demand. Two hundred copies would fit on that card, so the hardware question is not whether it fits but how many instances you stack per accelerator. The job is fill mask on case-sensitive English text: 355M parameters across 24 layers predict a masked token.
MIT is as light as a license gets: commercial use, modification and redistribution, with the copyright and permission notices kept in the package. Context is 514 tokens, so longer documents get chunked first. The 7.5 GB of files spans five formats, safetensors, ONNX, PyTorch, JAX and TensorFlow; pull the one your serving stack loads. Pretraining data was BookCorpus and Wikipedia, English only; the method is in arXiv:1907.11692.
Questions
Which is larger, deberta-v2-large-japanese-char-wwm or roberta-large?
roberta-large (355M parameters) is larger than deberta-v2-large-japanese-char-wwm (330M parameters), by the parameter counts their publishers report.
Which is cheaper to run, deberta-v2-large-japanese-char-wwm or roberta-large?
At 4-bit, deberta-v2-large-japanese-char-wwm fits on 1x MI300X from $1.85 an hour and roberta-large on 1x MI300X from $1.85 an hour, at the lowest on-demand prices the SAVRN Index lists.
Can I use deberta-v2-large-japanese-char-wwm commercially?
Yes. deberta-v2-large-japanese-char-wwm is released under Creative Commons Attribution-ShareAlike 4.0. CC BY-SA 4.0 permits sharing and adapting, including commercially, with credit to the creator, and requires adaptations to be released under the same license.
Can I use roberta-large commercially?
Yes. roberta-large is released under MIT License. The MIT License is a short permissive license. It permits commercial use, modification and redistribution, provided the copyright notice and permission notice are included.