MODEL CATALOGUE / HUB-DERIVED
Every checkpoint, sized exactly.
Leaderboards rank a model name. This catalogue addresses an artifact: a repository at a specific commit, with its parameter count and checkpoint bytes summed from the actual safetensors index rather than rounded off a press release.
Derived mechanically from the Hugging Face Hub API. Not a measurement, not a ranking, and not a claim that any runtime can load these weights.
WHY THIS IS DIFFERENT
A name is not an artifact.
The same model name can point at a BF16 release, an FP8 release, a community 4-bit requantisation and a fine-tune, each with different memory, different behaviour and a different licence. Every row here carries the commit SHA it describes, so two rows that disagree are telling you something true rather than contradicting each other.
Sizes are summed per dtype. A checkpoint that ships FP8 tensors is sized at one byte per parameter, not two — the difference is roughly 680 GB on a frontier sparse model, which is the difference between eight accelerators and sixteen.
Snapshot 2026-09-05 · regenerate with npm run ingest:hub · open the catalogue API
exact tensor sum repo — repository storage, used where weights are packed into wider containers and tensor arithmetic would overstate the size proj — parameters projected at the chosen precision, not a property of any published file
| Repository | Revision | Architecture | Parameters | Checkpoint | Experts | Licence | Downloads |
|---|---|---|---|---|---|---|---|
| QwenQwen3-0.6B | c1899de | qwen3 | 752M | 1.5 GB | — | apache-2.0 | 22.0M |
| trl-internal-testingtiny-Qwen2ForCausalLM-2.5 | 4b10ebe | qwen2 | 2M | 5 MB | — | Undeclared | 17.8M |
| openai-communitygpt2 | 607a30d | gpt2 | 137M | 548 MB | — | mit | 14.7M |
| QwenQwen3-8B | b968826 | qwen3 | 8.2B | 16.4 GB | — | apache-2.0 | 13.5M |
| QwenQwen2.5-7B-Instruct | a09a354 | qwen2 | 7.6B | 15.2 GB | — | apache-2.0 | 11.4M |
| nvidiaQwen3.6-35B-A3B-NVFP4 | 1355db6 | qwen3_5_moe | — | 21.4 GB↓ | — | apache-2.0 | 10.3M |
| QwenQwen2.5-1.5B-Instruct | 989aa79 | qwen2 | 1.5B | 3.1 GB | — | apache-2.0 | 7.4M |
| QwenQwen2.5-3B-Instruct | aa8e725 | qwen2 | 3.1B | 6.2 GB | — | other | 7.3M |
| QwenQwen3-Embedding-0.6B | 97b0c61 | qwen3 | 596M | 1.2 GB | — | apache-2.0 | 7.1M |
| farbodtavakkoliOTel-2.0-LLM-31B-IT | 522937f | gemma4 | 31.3B | 62.5 GB | — | apache-2.0 | 6.8M |
| openaigpt-oss-20b | 6cee5e8 | gpt_oss | 20.9B | 22.7 GB↓ | ? / 4 | apache-2.0 | 6.6M |
| QwenQwen3-4B | 1cfa9a7 | qwen3 | 4.0B | 8.0 GB | — | apache-2.0 | 6.3M |
| meta-llamaLlama-3.2-1B-Instructgated | 9213176 | llama | 1.2B | 2.5 GB | — | llama3.2 | 6.2M |
| QwenQwen2.5-0.5B-Instruct | 7ae5576 | qwen2 | 494M | 988 MB | — | apache-2.0 | 5.9M |
| meta-llamaLlama-3.1-8B-Instructgated | 0e9e39f | llama | 8.0B | 16.1 GB | — | llama3.1 | 5.7M |
| openaigpt-oss-120b | b5c939d | gpt_oss | 117B | 119.0 GB↓ | ? / 4 | apache-2.0 | 5.3M |
| QwenQwen3-32B | 9216db5 | qwen3 | 32.8B | 65.5 GB | — | apache-2.0 | 5.1M |
| dphndolphin-2.9.1-yi-1.5-34b | 0141cba | llama | 34.4B | 68.8 GB | — | apache-2.0 | 4.8M |
| deepseek-aiDeepSeek-V4-Flash-0731 | 7872f01 | deepseek_v4 | 304B | 166.9 GBrepo | ? / 6 | mit | 4.6M |
| QwenQwen-72B | b8e18ac | qwen | 72.3B | 144.6 GB | — | other | 3.9M |
| QwenQwen3-4B-Instruct-2507 | cdbee75 | qwen3 | 4.0B | 8.0 GB | — | apache-2.0 | 3.6M |
| QwenQwen3-1.7B | 70d244c | qwen3 | 2.0B | 4.1 GB | — | apache-2.0 | 3.5M |
| EleutherAIpythia-160m | 50f5173 | gpt_neox | 213M | 375 MB↓ | — | apache-2.0 | 3.5M |
| ornith-aiOrnith-1.0-35B | 5df2ed3 | qwen3_5_moe | — | 70.2 GB | — | mit | 3.0M |
| QwenQwen3-Embedding-4B | 5cf2132 | qwen3 | 4.0B | 8.0 GB | — | apache-2.0 | 2.8M |
| RadixArkKimi-K3-DSpark | 3c5bac3 | qwen3 | 2.2B | 4.5 GB | — | Undeclared | 2.7M |
| QwenQwen2.5-14B-Instruct | cf98f3b | qwen2 | 14.8B | 29.5 GB | — | apache-2.0 | 2.7M |
| HuggingFaceTBSmolLM2-135M | 93efa2f | llama | 135M | 269 MB | — | apache-2.0 | 2.6M |
| googlegemma-3-1b-itgated | dcc83ea | gemma3_text | 1000M | 2.0 GB | — | gemma | 2.6M |
| QwenQwen3-Reranker-4B | 22e6836 | qwen3 | 4.0B | 8.0 GB | — | apache-2.0 | 2.5M |
| QwenQwen2.5-Coder-7B-Instruct | c03e6d3 | qwen2 | 7.6B | 15.2 GB | — | apache-2.0 | 2.3M |
| QwenQwen3-Embedding-8B | 1d8ad4c | qwen3 | 7.6B | 15.1 GB | — | apache-2.0 | 2.2M |
| QwenQwen3-30B-A3B | ad44e77 | qwen3_moe | 30.5B | 61.1 GB | 128 / 8 | apache-2.0 | 2.2M |
| nvidiaNVIDIA-Nemotron-3-Nano-4B-BF16 | dfaf35d | nemotron_h | 4.0B | 7.9 GB | — | other | 2.2M |
| QwenQwen2.5-32B-Instruct | 5ede1c9 | qwen2 | 32.8B | 65.5 GB | — | apache-2.0 | 2.1M |
| distilbertdistilgpt2 | 2290a62 | gpt2 | 88M | 353 MB | — | apache-2.0 | 2.0M |
| QwenQwen3-4B-Base | 906bfd4 | qwen3 | 4.0B | 8.0 GB | — | apache-2.0 | 2.0M |
| zai-orgGLM-4.7-Flash | 7dd2089 | glm4_moe_lite | 31.2B | 62.4 GB | ? / 4 | mit | 1.9M |
| QwenQwen2.5-Coder-14B-Instruct | aedcc2d | qwen2 | 14.8B | 29.5 GB | — | apache-2.0 | 1.9M |
| QwenQwen3-1.7B-Base | ea980cb | qwen3 | 1.7B | 3.4 GB | — | apache-2.0 | 1.9M |
| deepseek-aiDeepSeek-V4-Flash | 60d8d70 | deepseek_v4 | 291B | 159.6 GBrepo | ? / 6 | mit | 1.8M |
| trl-internal-testingtiny-Qwen3ForCausalLM | 52b2e48 | qwen3 | 2M | 5 MB | — | Undeclared | 1.8M |
| QwenQwen3-14B | 40c0698 | qwen3 | 14.8B | 29.5 GB | — | apache-2.0 | 1.8M |
| QwenQwen2.5-Coder-32B-Instruct | 381fc96 | qwen2 | 32.8B | 65.5 GB | — | apache-2.0 | 1.7M |
| QwenQwen3-Coder-Next-FP8 | da6e2ed | qwen3_next | 79.7B | 80.4 GB↓ | 512 / 10 | apache-2.0 | 1.7M |
| QwenQwen2.5-0.5B | 060db64 | qwen2 | 494M | 988 MB | — | apache-2.0 | 1.7M |
| TinyLlamaTinyLlama-1.1B-Chat-v1.0 | fe8a4ea | llama | 1.1B | 2.2 GB | — | apache-2.0 | 1.7M |
| vikhyatkmoondream2 | 6b714b2 | moondream1 | 1.9B | 3.9 GB | — | apache-2.0 | 1.7M |
| deepseek-aiDeepSeek-V3.2 | a7e62ac | deepseek_v32 | 685B | 689.5 GB↓ | ? / 8 | mit | 1.6M |
| prism-mlBonsai-27B-mlx-1bit | ef22f23 | qwen3_5 | 1.7B | 5.1 GB | — | apache-2.0 | 1.6M |
| prism-mlTernary-Bonsai-27B-mlx-2bit | 70f75f3 | qwen3_5 | 27.4B | 8.5 GBrepo | — | apache-2.0 | 1.6M |
| QuantTrioQwen3-VL-30B-A3B-Instruct-AWQ | a5ea107 | qwen3_vl_moe | 31.1B | 17.9 GBrepo | — | apache-2.0 | 1.5M |
| meta-llamaLlama-3.2-3B-Instructgated | 0cb88a4 | llama | 3.2B | 6.4 GB | — | llama3.2 | 1.4M |
| meta-llamaMeta-Llama-3-8B-Instructgated | 8afb486 | llama | 8.0B | 16.1 GB | — | llama3 | 1.4M |
| appleOpenELM-1_1B-Instruct | effd796 | openelm | 1.1B | 2.2 GB | — | apple-amlr | 1.4M |
| HuggingFaceTBSmolLM2-135M-Instruct | 12fd25f | llama | 135M | 269 MB | — | apache-2.0 | 1.4M |
| microsoftphi-2 | 810d367 | phi | 2.8B | 5.6 GB | — | mit | 1.4M |
| zai-orgGLM-5.2-FP8 | f33c6dc | glm_moe_dsa | 753B | 755.4 GB↓ | ? / 8 | mit | 1.3M |
| googlegemma-3-270mgated | 9b0cfec | gemma3_text | 268M | 536 MB | — | gemma | 1.3M |
| meta-llamaLlama-3.2-1Bgated | 4e20de3 | llama | 1.2B | 2.5 GB | — | llama3.2 | 1.3M |