Up to date
Nanbeige Apache 2.0 out 8 months ago
3.93B parameters, text in, text out, 256K tokens of context. 1 provider serves it.
Verified 2026-10-25 Rank #658 in LLMs · score 73 method v1 Sources 3 Downloads 30d 3.7k · #561 of 619 tracked
Changed this week: 1 new quantized build on the Hub (62 in all).2026-10-24
Picked by rule from 1 tracked piece: the lab's own post, the most shared, a field report, the latest.
Live hosted endpoints, cheapest first, USD per million tokens. Ties break on output price, then provider name.
| # | Cheapest live row | In /1M | Out /1M | Context | Quant | Seen |
|---|---|---|---|---|---|---|
| 1 | Featherless | — | — | — | — | Oct 25 |
Hugging Face does not say who downloads a model or with what. These are the proxies: API providers listing it, quantized builds, fine-tunes/adapters/merges, and Spaces that load it.
Served by 1 API provider; heavily run locally (62 quantized builds).
Top Spaces running it: victor/nanbeige ♥ 35 · GenAISecurityProject/OWASP-AIBOM-Generator ♥ 30 · yaris664/Nanbeige4.1-3B-space ♥ 1
Weights only, an estimate: leave room for the KV cache. Method
About 2.4 GB at INT4, 4.7 GB at FP8, 9.4 GB at BF16: weights only, so leave room for the KV cache.
Estimate params × bytes per param × 1.2 · how we estimate
Weights, every dated article, and repo commits on one rail.
Teal is the weights (2026-02-10), amber triangles are 1 dated article (each links to it), grey ticks are repo commits.
100 derivatives on the Hub. Which build to run, by runtime: the most downloaded in each class.
From the Hugging Face model tree; quantized builds in teal, fine-tunes, merges and adapters in greys.
Everything else, folded: the full tables, every source.
Labels are evidence, not verdicts: measured a third-party leaderboard row for this model; author-reported a number the model author published in the card or eval files; structural a deterministic signal from the name, config or chat template, or a decision-model estimate. Each chip links to its source. All use cases.
Listed in LLMs
| Pipeline tag | text-generation |
|---|---|
| Library | transformers |
| Architectures | LlamaForCausalLM |
| Model type | llama |
| Dtype | bfloat16 |
| Params by dtype | BF16: 3,933,637,120 |
| Size band | M |
| MoE | no |
| Context length | 262,144 (config.json max_position_embeddings) |
| Languages | en, zh |
| Papers | arXiv:2602.13367 |
| License | Apache 2.0 (apache-2.0) |
| Hub snapshot | 2026-10-25 |
vLLM: supported · LlamaForCausalLM · 7050ef881ff0835f6890a47db1859a24c65481d6SGLang: supported · LlamaForCausalLM · 773cb938fe9484a8f7e7e46fb1b1fe96412c59ed
vLLM registry (7050ef881ff0835f6890a47db1859a24c65481d6) · SGLang registry (773cb938fe9484a8f7e7e46fb1b1fe96412c59ed) · Matched by architecture class name against each project's registry on main; 'supported' means the architecture is registered, not that this checkpoint was tested.
Weights-only memory for Nanbeige4.1-3B: params × bytes/param × 1.2 overhead. KV cache and activations are extra. How this is computed.
| BF16 (2 bytes/param) | 9.4 GB · Single GPU (<24 GB) |
|---|---|
| FP8 (1 byte/param) | 4.7 GB · Single GPU (<24 GB) |
| INT4 (0.5 bytes/param) | 2.4 GB · Edge (<4 GB) |
Weights only (params × bytes/param × 1.2 overhead); KV cache and activations extra.
Base model: Nanbeige/Nanbeige4-3B-Base
quantized: 62finetune: 28adapter: 8merge: 2
| Variant | Relation | Quant | Params | Band | License | Author | Likes | Downloads |
|---|---|---|---|---|---|---|---|---|
| Edge-Quant/Nanbeige4.1-3B-Q4_K_M-GGUF | quantized | GGUF | — | — | apache-2.0 | Edge-Quant | 40 | 130.4k |
| Edge-Quant/Nanbeige4.1-3B-Q8_0-GGUF | quantized | Q8_0 | — | — | apache-2.0 | Edge-Quant | 6 | 123.5k |
| DevQuasar/Nanbeige.Nanbeige4.1-3B-GGUF | quantized | GGUF | — | — | — | DevQuasar | 6 | 94.1k |
| cyankiwi/Nanbeige4.1-3B-AWQ-8bit | quantized | AWQ | 3.93B | M | apache-2.0 | cyankiwi | 2 | 49.9k |
| mradermacher/Nanbeige4.1-3B-GGUF | quantized | GGUF | — | — | apache-2.0 | mradermacher | 38 | 35.5k |
| tantk/Nanbeige4.1-3B-GGUF | quantized | GGUF | — | — | apache-2.0 | tantk | 5 | 14.8k |
| mradermacher/Nanbeige4.1-3B-i1-GGUF | quantized | GGUF | — | — | apache-2.0 | mradermacher | 6 | 12.7k |
| RoroMachine14/nanbeige-4.1-3b-deepseek-distilled | quantized | GGUF | 3.93B | M | apache-2.0 | RoroMachine14 | 4 | 9.7k |
| kaitchup/Nanbeige4.1-3B-autoround-W4A16 | quantized | -W4A16 | 1.40B | S | apache-2.0 | kaitchup | 1 | 8k |
| mlx-community/Nanbeige4.1-3B-8bit | quantized | -8bit | 3.93B | M | apache-2.0 | mlx-community | 15 | 7k |
| heretic-org/Nanbeige4.1-3B-heretic | finetune | — | 3.93B | M | apache-2.0 | heretic-org | 52 | 4.7k |
| Open4bits/Nanbeige4.1-3B-GGUF | quantized | GGUF | — | — | apache-2.0 | Open4bits | 0 | 6.7k |
| deltakitsune/Nanbeige-4.1-Python-DeepThink-3B | adapter | GGUF | 3.93B | M | apache-2.0 | deltakitsune | 8 | 5.4k |
| seanbailey518/Nanbeige4.1-3B-GGUF | quantized | GGUF | — | — | other | seanbailey518 | 0 | 5.1k |
| prithivMLmods/Nanbeige4.1-3B-f32-GGUF | quantized | GGUF | — | — | llama3 | prithivMLmods | 4 | 4.7k |
| onnx-community/Nanbeige4.1-3B-ONNX | quantized | — | — | — | apache-2.0 | onnx-community | 4 | 2.1k |
| Akicou/Nanbeige4.1-3B-GGUF | quantized | GGUF | — | — | — | Akicou | 0 | 2.3k |
| cyankiwi/Nanbeige4.1-3B-AWQ-4bit | quantized | AWQ | 3.93B | M | apache-2.0 | cyankiwi | 1 | 2.2k |
| SimplySara/Nanbeige4.1-3B-GGUF | quantized | GGUF | — | — | apache-2.0 | SimplySara | 1 | 2.1k |
| SimplySara/Nanbeige4.1-3B-i1-GGUF | quantized | GGUF | — | — | apache-2.0 | SimplySara | 1 | 2k |
| mlx-community/Nanbeige4.1-3B-bf16 | finetune | -bf16 | 3.93B | M | apache-2.0 | mlx-community | 3 | 1.5k |
| megabytes/Nanbeige4.1-3B-heretic | finetune | — | 3.93B | M | apache-2.0 | megabytes | 17 | 661 |
| andrevp/Nanbeige4.1-3B-MLX-4bit | quantized | MLX | 3.93B | M | apache-2.0 | andrevp | 4 | 1.2k |
| McG-221/Nanbeige4.1-3B-mlx-8Bit | quantized | mlx | 3.93B | M | apache-2.0 | McG-221 | 0 | 1.4k |
| ThalisAI/Nanbeige4.1-3B-heretic | quantized | GGUF | — | — | other | ThalisAI | 0 | 1.3k |
| DhruvalLabs/Nanbeige4.1-3B-GGUF | quantized | GGUF | — | — | apache-2.0 | DhruvalLabs | 0 | 1.1k |
| NikolayKozloff/Nanbeige4.1-3B-Q8_0-GGUF | quantized | Q8_0 | — | — | apache-2.0 | NikolayKozloff | 1 | 917 |
| kaitchup/Nanbeige4.1-3B-FP8-Dynamic | quantized | -FP8 | 3.93B | M | apache-2.0 | kaitchup | 0 | 917 |
| litert-community/Nanbeige4.1-3B | quantized | — | — | — | apache-2.0 | litert-community | 2 | 786 |
| mlx-community/Nanbeige4.1-3B-4bit | quantized | -4bit | 3.93B | M | apache-2.0 | mlx-community | 1 | 815 |
| Edge-Quant/Nanbeige4.1-3B-Q6_K-GGUF | quantized | Q6_K | — | — | apache-2.0 | Edge-Quant | 2 | 691 |
| NeshVerse/Uncensored_Nanbeige-4.1-3B | finetune | — | 3.93B | M | apache-2.0 | NeshVerse | 3 | 457 |
| TurkishCodeMan/Nanbeige4.1-3B-Gmail-Tool-Use | finetune | — | 3.93B | M | apache-2.0 | TurkishCodeMan | 2 | 475 |
| emee-ai/Nanbeige4.1-3B-bf16 | finetune | -bf16 | 3.93B | M | apache-2.0 | emee-ai | 0 | 553 |
| alexlink/Nanbeige4.1-3B-GGUF | quantized | GGUF | — | — | — | alexlink | 0 | 544 |
| SkyAsl/Nanbeige4.1-VLM | finetune | — | 3.94B | M | apache-2.0 | SkyAsl | 1 | 460 |
| liodon-ai/Nanbeige4.1-3B-imatrix-GGUF | quantized | imatrix | — | — | other | liodon-ai | 0 | 499 |
| RepublicOfKorokke/Nanbeige4.1-3B-mlx-mxfp4 | quantized | mlx | 3.93B | M | apache-2.0 | RepublicOfKorokke | 1 | 426 |
| amoughnieh/Nanbeige4.1-3B-EE-Reasoning-SFT | adapter | GGUF | — | — | mit | amoughnieh | 1 | 418 |
| NeshVerse/Ultra_Flash_Financial_SFT_Nanbeige_4.1-3B | merge | — | 4.03B | M | apache-2.0 | NeshVerse | 4 | 233 |
| Date | Commit | SHA |
|---|---|---|
| 2026-03-25 | Update tokenizer_config.json | 9c4555c |
| 2026-02-26 | Add syntax highlight in python markdown code snippets (#34) | 6f3b2c3 |
| 2026-02-26 | increase the max_new_tokens in the demo (#33) | 971753c |
| 2026-02-22 | Update README.md | 2d78edc |
| 2026-02-19 | Update README.md | 7e64fe9 |
| 2026-02-17 | Update README.md | 988060c |
| 2026-02-13 | Update README.md | b32903a |
| 2026-02-13 | Upload Nanbeige4.1-3B-Report.pdf | 57b5051 |
| 2026-02-11 | Update config.json | 0133423 |
| 2026-02-11 | Upload model_performance_comparison.png | 5fa91af |
| 2026-02-11 | Delete figures/model_performance_comparison.png | b7eedac |
| 2026-02-11 | Update README.md | da13918 |
| 2026-02-11 | Upload model_performance_comparison.png | 28f4b2e |
| 2026-02-11 | Update README.md | d02c34c |
| 2026-02-11 | Delete figures/1 | 8109f12 |
| 2026-02-11 | Upload nbg.png | 9c1da7f |
| 2026-02-11 | Create figures/1 | 810318d |
| 2026-02-11 | Upload tokenizer_config_search.json | 253dcb4 |
| 2026-02-10 | Upload 2 files | 665db78 |
| 2026-02-10 | Upload 2 files | 1f158ff |
| 2026-02-10 | Upload 6 files | 346a63d |
| 2026-02-10 | initial commit | 7d222b4 |
Hosted endpoints observed for Nanbeige4.1-3B, sorted by input price (USD per million tokens). Each row names its observer and date; prices change daily, so check the source before quoting.
| Provider | Model ref | Context | Quant | $/M in | $/M out | tok/s | TTFT | Tools | Status | Observed |
|---|---|---|---|---|---|---|---|---|---|---|
| Featherless hf | Nanbeige/Nanbeige4.1-3B | — | — | — | — | — | — | — | live | 2026-10-25 |
Ranked by source signals (HN, Lobsters, authority domains, official posts) and recency.
arXiv links found in the model card and its tags. "Peer-reviewed" is shown only with evidence: a published venue in OpenAlex (measured) or the authors' own journal reference on arXiv (reported).
Hub snapshot: 2026-10-25 · HF API: 2026-10-24 21:42 UTC · Providers: 2026-10-24 21:44 UTC · Articles: 2026-10-24 21:46 UTC · Summary: 2026-10-07 15:14 UTC · Last verified 2026-10-25
Email alerts are not open yet; the feeds above are the same data. See feeds for the item schema.
