Up to date
bigscience BLOOM RAIL 1.0 out 4 years ago
176B parameters, text in, text out.
Verified 2026-10-25 Rank #195 in LLMs · score 89.4 method v1 Sources 3 Downloads 30d 11.8k · #499 of 619 tracked
Picked by rule from 9 tracked pieces: the lab's own post, the most shared, a field report, the latest.
The 1 pick above, plus these 8. Top is the radar's score: discussion, source authority and the lab's own posts, fading with age.
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.
No API provider serves it, so its downloads are people running it themselves: 2 quantized builds and 100 Spaces.
Top Spaces running it: uwnlp/guanaco-playground-tgi ♥ 507 · olivierdehaene/chat-llm-streaming ♥ 336 · huggingface/bloom_demo ♥ 286 · tomg-group-umd/lm-watermarking ♥ 100
Weights only, an estimate: leave room for the KV cache. Method
About 106 GB at INT4, 212 GB at FP8, 423 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 (2022-05-19), amber triangles are 5 dated articles (each links to it), grey ticks are repo commits.
7 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.
BLOOM is a 176B parameter decoder-only autoregressive language model developed by bigscience.[1][2] It supports training data in 46 natural languages and 13 programming languages.[3]
Generated 2026-10-07 15:04 UTC by parasail-gemma-4-26b-a4b-it; each numeral points at the fact it rests on. Not a verdict.
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 | BloomForCausalLM |
| Model type | bloom |
| Params by dtype | BF16: 176,247,271,424 |
| Size band | XXL |
| MoE | no |
| Languages | ak, ar, as, bm, bn, ca, code, en, es, eu, fon, fr, gu, hi, id, ig, ki, kn, lg, ln, ml, mr, ne, nso, ny, or, pa, pt, rn, rw, sn, st, sw, ta, te, tn, ts, tum, tw, ur, vi, wo, xh, yo, zh, zu |
| Papers | arXiv:2211.05100, arXiv:1909.08053, arXiv:2110.02861, arXiv:2108.12409 |
| License | BLOOM RAIL 1.0 (bigscience-bloom-rail-1.0) |
| Hub snapshot | 2026-10-25 |
vLLM: supported · BloomForCausalLM · 7050ef881ff0835f6890a47db1859a24c65481d6SGLang: not in registry · 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 bloom: params × bytes/param × 1.2 overhead. KV cache and activations are extra. How this is computed.
| BF16 (2 bytes/param) | 423 GB · Multi-GPU node (<640 GB) |
|---|---|
| FP8 (1 byte/param) | 212 GB · Multi-GPU node (<640 GB) |
| INT4 (0.5 bytes/param) | 106 GB · Multi-GPU node (<640 GB) |
Weights only (params × bytes/param × 1.2 overhead); KV cache and activations extra.
quantized: 2finetune: 3merge: 2
| Variant | Relation | Quant | Params | Band | License | Author | Likes | Downloads |
|---|---|---|---|---|---|---|---|---|
| DevQuasar-13/bigscience.bloom-GGUF | quantized | GGUF | — | — | — | DevQuasar-13 | 1 | 2.7k |
| mradermacher/bloom-i1-GGUF | quantized | GGUF | — | — | bigscience-bloom-rail-1.0 | mradermacher | 3 | 1.4k |
| 12sebas/new_jar | finetune | — | — | — | — | 12sebas | 0 | 436 |
| Or4cl3-1/SAM-Gemini-BLOOM-OPT-Gopher-Megatron-slerp | merge | — | — | — | — | Or4cl3-1 | 0 | 0 |
| Or4cl3-1/code-slerp | merge | — | — | — | — | Or4cl3-1 | 0 | 0 |
| mradermacher/bloom-GGUF | finetune | GGUF | — | — | bigscience-bloom-rail-1.0 | mradermacher | 0 | 0 |
| MLap/bloom1.7-lora-sentiment-analysis-classification | finetune | — | — | — | mit | MLap | 0 | 0 |
| Date | Commit | SHA |
|---|---|---|
| 2026-07-29 | Fixed CO2 metadata in README.md (#293) | 7f10a99 |
| 2023-07-28 | Fix asset URL | 053d9cd |
| 2023-04-05 | Upload model_00072-of-00072.safetensors (#230) | 87cbfc8 |
| 2023-03-14 | clarify | 476113c |
| 2023-03-14 | add a link to https://huggingface.co/bigscience/bloom-intermediate/ | 0a8c4dc |
| 2023-02-10 | Fix architecture (#189) | ea51bbb |
| 2023-02-10 | Fix architecture (#188) | dede1ba |
| 2023-02-02 | Fix typo in Examples (#181) | b366374 |
| 2022-11-30 | Update README.md | f0e3b92 |
| 2022-11-30 | Update README.md (#143) | 0a1bf22 |
| 2022-11-18 | Add safetensors index (#142) | 4bf48c6 |
| 2022-11-14 | Add carbon emissions metadata (#140) | 20ac6a9 |
| 2022-10-21 | Adding `safetensors` variant of this model (#121) | 4d8e28c |
| 2022-09-29 | Fix dataset name | 4ab0472 |
| 2022-09-08 | Add exact param counts (#106) | 0f24f91 |
| 2022-09-08 | Fix mathematical reasoning (#103) | c3a35dc |
| 2022-09-01 | Update default template for translation to spanish | 11a7c1b |
| 2022-09-01 | Improve default translation prompts (#102) | 8540347 |
| 2022-08-31 | Fix #98 | 874fa44 |
| 2022-08-31 | Remove incorrect evaluation results. (#98) | 3f4b529 |
| 2022-08-16 | Correct HumanEval scores (#79) | aee8c40 |
| 2022-07-29 | Widget multiline examples correct behaviour (#67) | d9bf58e |
| 2022-07-25 | cross link to meg-ds checkpoint | 9e27a90 |
| 2022-07-22 | Add Lingala (#60) | 2a3d62e |
| 2022-07-21 | make the release date human-friendly | 483e9e2 |
| 2022-07-20 | Update README.md (#57) | 4051715 |
| 2022-07-18 | Fix the tokenizer class (#53) | e1d88ec |
| 2022-07-18 | Update README.md | 373a962 |
| 2022-07-16 | Fix language name | 7fc8195 |
| 2022-07-13 | Update README.md (#44) | e965be3 |
Ranked by source signals (HN, Lobsters, authority domains, official posts) and recency. Summaries are generated from a stored copy of the article text; every sentence cites a span of that text, and nothing is shown that failed the citation check.
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 21:37 UTC · Providers: 2026-10-24 21:44 UTC · Articles: 2026-10-24 21:45 UTC · Summary: 2026-10-07 15:04 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.
