numbat is an open-weight model from Burner22. It has 131,072-token context. Its published files total 1.1 TB.
~15T param, 114 layers, hybrid GDN/MLA, MoE (1,224 experts, top-48 + 1 shared), NVFP4. Internal codename numbat, Anthropic's next Mythos-class training run. numbat is the codename referenced in the Claude Code source leak (Mar 2026).
Parameters—
Context131,072
Weights1.1 TB
License—
AccessOpen weights
Monthly Downloads—
Model Card
~15T param, 114 layers, hybrid GDN/MLA, MoE (1,224 experts, top-48 + 1 shared), NVFP4. Internal codename numbat, Anthropic's next Mythos-class training run. numbat is the codename referenced in the Claude Code source leak (Mar 2026). This is a checkpoint from that training run, pulled mid-pretraining. The 842 shards total ~8.42 TB of tensor data. Each shard is safetensors format, under 10.6 GB per file. configurationnumbat.py and modelingnumbat.py are included: a complete reference implementation of the checkpoint. Its module tree has an exact slot for every tensor in every shard, so the full checkpoint loads with loadstatedict(strict=True) (and, since the classes derive from…
Excerpt from the card by Burner22.
Configuration
- Architecture
- NumbatForCausalLM
- Context length (tokens)
- 131,072
- Layers
- 114
- Hidden size
- 12,288
- Feed-forward size
- 24,576
- Vocabulary size
- 131,072
- Model type
- numbat
- Quantization
- modelopt
Identity and Version
- Repository
- burner22/numbat
- Publisher
- Burner22
- Task
- Not stated by the source
- Modality
- Other
- Library
- Not stated by the source
- Parameters
- Not stated by the source
- Languages
- Not stated by the source
- Revision
- f20c45411496b7330ce7a3092cb13b8a1e562ace
- First published
- 2026-10-18
- Last updated
- 2026-10-18
Files and Weights
121 files, 1.1 TB in total. The weights are 109 files totalling 1.1 TB in bin, safetensors.
Weights109 files · 1.1 TB
Configuration7 files · 145.3 MB
Tokenizer2 files · 8.9 MB
Documentation1 file · 3.6 KB
Other1 file · 66.4 KB
Repository1 file · 1.6 KB
Every file
| File | Type | Size | SHA-256 |
|---|---|---|---|
| model-00001-of-00842.safetensors | Weights | 10.1 GB | 871acd5f456a |
| model-00002-of-00842.safetensors | Weights | 10.1 GB | f5b672e9d0e0 |
| model-00003-of-00842.safetensors | Weights | 10.1 GB | 77c400c5d87c |
| model-00004-of-00842.safetensors | Weights | 10.1 GB | d72440bb3f2c |
| model-00005-of-00842.safetensors | Weights | 10.1 GB | 11e99aa3c18c |
| model-00006-of-00842.safetensors | Weights | 10.1 GB | 74d032a33704 |
| model-00007-of-00842.safetensors | Weights | 10.1 GB | 9810e158eadf |
| model-00008-of-00842.safetensors | Weights | 10.1 GB | 6cb319939ae2 |
| model-00009-of-00842.safetensors | Weights | 10.1 GB | dc113a44723b |
| model-00010-of-00842.safetensors | Weights | 10.1 GB | 83027d515746 |
| model-00011-of-00842.safetensors | Weights | 10.1 GB | 4b602c75da73 |
| model-00012-of-00842.safetensors | Weights | 10.1 GB | 2daf45cd3a7d |
| model-00013-of-00842.safetensors | Weights | 10.1 GB | 0a7cbf01fea0 |
| model-00014-of-00842.safetensors | Weights | 10.1 GB | 1f0e1b413c44 |
| model-00015-of-00842.safetensors | Weights | 10.1 GB | f76979ddbb10 |
| model-00016-of-00842.safetensors | Weights | 10.1 GB | 1b868e1a8f9a |
| model-00017-of-00842.safetensors | Weights | 10.1 GB | e07db1677575 |
| model-00018-of-00842.safetensors | Weights | 10.1 GB | cd7133ec3cf5 |
| model-00019-of-00842.safetensors | Weights | 10.1 GB | 82b4f2bab45a |
| model-00141-of-00842.safetensors | Weights | 10.1 GB | 1c45a550f611 |
| model-00142-of-00842.safetensors | Weights | 10.1 GB | 8df83ae85acb |
| model-00143-of-00842.safetensors | Weights | 10.1 GB | 74ff631a7b80 |
| model-00144-of-00842.safetensors | Weights | 10.1 GB | 52b1b3c3c740 |
| model-00145-of-00842.safetensors | Weights | 10.1 GB | 17a9bc0b33c0 |
| model-00146-of-00842.safetensors | Weights | 10.1 GB | 0e7eece96829 |
| model-00147-of-00842.safetensors | Weights | 10.1 GB | 8095e002e327 |
| model-00148-of-00842.safetensors | Weights | 10.1 GB | 803c968bdc50 |
| model-00149-of-00842.safetensors | Weights | 10.1 GB | 17495cbc8afb |
| model-00150-of-00842.safetensors | Weights | 10.1 GB | a55b32aa8031 |
| model-00151-of-00842.safetensors | Weights | 10.1 GB | cbe47aa893b0 |
| model-00152-of-00842.safetensors | Weights | 10.1 GB | 25257dab3897 |
| model-00153-of-00842.safetensors | Weights | 10.1 GB | ddbd21be99d8 |
| model-00154-of-00842.safetensors | Weights | 10.1 GB | dca1d706b6d9 |
| model-00155-of-00842.safetensors | Weights | 10.1 GB | 367dcf5d1fb8 |
| model-00156-of-00842.safetensors | Weights | 10.1 GB | f66ac4f40e25 |
| model-00281-of-00842.safetensors | Weights | 10.1 GB | 1e97de498271 |
| model-00282-of-00842.safetensors | Weights | 10.1 GB | 487367e8b02c |
| model-00283-of-00842.safetensors | Weights | 10.1 GB | e8bcc3360b90 |
| model-00284-of-00842.safetensors | Weights | 10.1 GB | 4553eb0556d8 |
| model-00285-of-00842.safetensors | Weights | 10.1 GB | e586294cec41 |
| model-00286-of-00842.safetensors | Weights | 10.1 GB | da0e7b5bcc0a |
| model-00287-of-00842.safetensors | Weights | 10.1 GB | 1f5c7a46419a |
| model-00288-of-00842.safetensors | Weights | 10.1 GB | 6c56e9a996d7 |
| model-00289-of-00842.safetensors | Weights | 10.1 GB | 7fb58fdf9405 |
| model-00290-of-00842.safetensors | Weights | 10.1 GB | 19841f413574 |
| model-00291-of-00842.safetensors | Weights | 10.1 GB | a7a901720021 |
| model-00292-of-00842.safetensors | Weights | 10.1 GB | 408356351e17 |
| model-00293-of-00842.safetensors | Weights | 10.1 GB | fcff7359cd3c |
| model-00294-of-00842.safetensors | Weights | 10.1 GB | 0f74adabcd88 |
| model-00295-of-00842.safetensors | Weights | 10.1 GB | 2f9b05745dc0 |
| model-00296-of-00842.safetensors | Weights | 10.1 GB | 29bcf4b3f05a |
| model-00421-of-00842.safetensors | Weights | 10.1 GB | 6802280e6ffd |
| model-00422-of-00842.safetensors | Weights | 10.1 GB | a076ffa59ee2 |
| model-00423-of-00842.safetensors | Weights | 10.1 GB | a90277363da0 |
| model-00424-of-00842.safetensors | Weights | 10.1 GB | c03d46bbba0e |
| model-00425-of-00842.safetensors | Weights | 10.1 GB | adb55d9bdeae |
| model-00426-of-00842.safetensors | Weights | 10.1 GB | 199f34799ea8 |
| model-00427-of-00842.safetensors | Weights | 10.1 GB | 55cb78b59abf |
| model-00428-of-00842.safetensors | Weights | 10.1 GB | 139833b77dcd |
| model-00429-of-00842.safetensors | Weights | 10.1 GB | 1f1a9d05e88d |
| model-00430-of-00842.safetensors | Weights | 10.1 GB | 60114539e06b |
| model-00431-of-00842.safetensors | Weights | 10.1 GB | 50b872fd407d |
| model-00432-of-00842.safetensors | Weights | 10.1 GB | b8b61b2d1c37 |
| model-00433-of-00842.safetensors | Weights | 10.1 GB | 5ebb5ecf4943 |
| model-00434-of-00842.safetensors | Weights | 10.1 GB | 8a2d3181562d |
| model-00435-of-00842.safetensors | Weights | 10.1 GB | 7e331c673f16 |
| model-00436-of-00842.safetensors | Weights | 10.1 GB | 428e123a04e5 |
| model-00437-of-00842.safetensors | Weights | 10.1 GB | dcc7161faa2e |
| model-00438-of-00842.safetensors | Weights | 10.1 GB | 041b550e2227 |
| model-00439-of-00842.safetensors | Weights | 10.1 GB | 7a4772e2e305 |
| model-00440-of-00842.safetensors | Weights | 10.1 GB | db5ccbd7368c |
| model-00441-of-00842.safetensors | Weights | 10.1 GB | 53e598dd0c97 |
| model-00561-of-00842.safetensors | Weights | 10.1 GB | 3a043e17e09e |
| model-00562-of-00842.safetensors | Weights | 10.1 GB | f8b9347b76e9 |
| model-00563-of-00842.safetensors | Weights | 10.1 GB | 155b93210b00 |
| model-00564-of-00842.safetensors | Weights | 10.1 GB | 344459cd0bd3 |
| model-00565-of-00842.safetensors | Weights | 10.1 GB | da3ccde6e124 |
| model-00566-of-00842.safetensors | Weights | 10.1 GB | 719651738653 |
| model-00567-of-00842.safetensors | Weights | 10.1 GB | 861964953a17 |
| model-00568-of-00842.safetensors | Weights | 10.1 GB | bc62792d6427 |
| model-00569-of-00842.safetensors | Weights | 10.1 GB | dfa9e3ad5c22 |
| model-00570-of-00842.safetensors | Weights | 10.1 GB | d9a23a9d72b1 |
| model-00571-of-00842.safetensors | Weights | 10.1 GB | d3df93218028 |
| model-00572-of-00842.safetensors | Weights | 10.1 GB | 230f045ac4c1 |
| model-00573-of-00842.safetensors | Weights | 10.1 GB | 106b97d6ad4d |
| model-00574-of-00842.safetensors | Weights | 10.1 GB | 77fd20697b95 |
| model-00575-of-00842.safetensors | Weights | 10.1 GB | 48d7c738cb79 |
| model-00576-of-00842.safetensors | Weights | 10.1 GB | e869367ccda9 |
| model-00577-of-00842.safetensors | Weights | 10.1 GB | 566fd57a6557 |
| model-00578-of-00842.safetensors | Weights | 10.1 GB | d29cb0e853c0 |
| model-00579-of-00842.safetensors | Weights | 10.1 GB | 29991afa4540 |
| model-00580-of-00842.safetensors | Weights | 10.1 GB | b621cbc243cf |
| model-00701-of-00842.safetensors | Weights | 10.1 GB | 7dc5453bbcc1 |
| model-00702-of-00842.safetensors | Weights | 10.1 GB | d8fcc5185f98 |
| model-00703-of-00842.safetensors | Weights | 10.1 GB | 708135c49070 |
| model-00704-of-00842.safetensors | Weights | 10.1 GB | ee823d712ef9 |
| model-00705-of-00842.safetensors | Weights | 10.1 GB | b7aa5bd5cc5e |
| model-00706-of-00842.safetensors | Weights | 10.1 GB | b0e81ffd3490 |
| model-00707-of-00842.safetensors | Weights | 10.1 GB | c03243f01e7c |
| model-00708-of-00842.safetensors | Weights | 10.1 GB | c1d52b8e6e04 |
| model-00709-of-00842.safetensors | Weights | 10.1 GB | b08873713b0f |
| model-00710-of-00842.safetensors | Weights | 10.1 GB | f6eb0ba6a511 |
| model-00711-of-00842.safetensors | Weights | 10.1 GB | d628c20bc59f |
| model-00712-of-00842.safetensors | Weights | 10.1 GB | f33e18b03986 |
| model-00713-of-00842.safetensors | Weights | 10.1 GB | ffd6a7ce2a2e |
| model-00714-of-00842.safetensors | Weights | 10.1 GB | 5bc8506f8d11 |
| model-00715-of-00842.safetensors | Weights | 10.1 GB | c9cdec25154b |
| model-00716-of-00842.safetensors | Weights | 10.1 GB | f506e56c12a3 |
| test-upload.bin | Weights | 13 B | a7639b92d9b5 |
| ckpt_metadata.json | Configuration | 810 B | — |
| config.json | Configuration | 4.2 KB | — |
| configuration_numbat.py | Configuration | 3.2 KB | — |
| generation_config.json | Configuration | 230 B | — |
| hf_quant_config.json | Configuration | 915 B | — |
| model.safetensors.index.json | Configuration | 145.3 MB | 97bba057c37e |
| modeling_numbat.py | Configuration | 22.3 KB | — |
| README.md | Documentation | 3.6 KB | — |
| training_log_tail.txt | Other | 66.4 KB | — |
| .gitattributes | Repository | 1.6 KB | — |
| tokenizer.json | Tokenizer | 8.9 MB | — |
| tokenizer_config.json | Tokenizer | 485 B | — |
License and Download
- License
- Not stated by the source
- Access
- Open weights, no gate
- Download size
- 1.1 TB
Download from Burner22
Released by Burner22 through its official repository on Hugging Face.
Memory Requirements
| Precision | Weights in memory |
|---|---|
| As published | 1.1 TB |
Weights only, from the published parameter count; the key-value cache and runtime add to this.
Questions About numbat
What is numbat's context length?
131,072 tokens, from the maximum position embeddings in its published configuration.