Model Follower.

Up to date

Rebuilt
2026-10-25 06:54 UTC
Release radar
117 on watch
Hub snapshot
2026-10-25
Re-checked
daily · every number sourced and dated

Models › Llama-3.1 family

Llama-3.1-405B-Instruct

Hugging Face ↗ Follow changes

meta-llama Llama 3.1 Community License out 2 years ago

406B parameters, text in, text out.

Verified 2026-10-25 Rank #371 in LLMs · score 80.5 method v1 Sources 3 Downloads 30d 9.4k · #512 of 619 tracked

Can I run it?244GB at INT4Estimateneeds a multi-GPU nodemethod
Can I use it?Llama 3.1 Community License'Built with Llama' attribution; >700M MAU clauselicence file2026-10-24
Do people use it?9.4kdownloads, 30 days#512 of 619 we track · below the median+0 likes, downloads −5% this week604 likesHugging Face2026-10-24

Worth reading

Picked by rule from 25 tracked pieces: the lab's own post, the most shared, a field report, the latest.

25 best of 25 tracked

The 2 picks above, plus these 23. Top is the radar's score: discussion, source authority and the lab's own posts, fading with age.

  1. CodeShow HN: Llama 3.1 70B on a single RTX 3090 via NVMe-to-GPU bypassing the CPUgithub.com · 2026-02-21
  2. ArticleLlama 3.1llama.meta.com · 2024-07-23 · 437 HN pts · official
  3. Codeg1: Using Llama-3.1 70B on Groq to create o1-like reasoning chainsgithub.com · 2024-09-15
  4. CodeLlama 3.1 Omni Modelgithub.com · 2024-09-18
  5. CodeShow HN: Tune LLaMa3.1 on Google Cloud TPUsgithub.com · 2024-09-11 · 189 HN pts
  6. CodeLlama 3.1 in Cgithub.com · 2024-07-24 · 212 HN pts
  7. CodeLongwriter – Increase llama3.1 output to 10k wordsgithub.com · 2024-10-07 · 154 HN pts
  8. ArticleHermes 3: The First Fine-Tuned Llama 3.1 405B Modellambdalabs.com · 2024-08-15 · 146 HN pts
  9. ArticleLlama 3.1: Our most capable models to dateai.meta.com · 2024-07-23 · 27 HN pts · official
  10. DiscussionCerebras launches inference for Llama 3.1; benchmarked at 1846 tokens/s on 8Btwitter.com · 2024-08-27 · 95 HN pts
  11. ArticleWhy Llama 3.1 is Importanttimkellogg.me · 2024-07-23 · 56 HN pts
  12. ArticleNvidia presents Llama-Mesh: Generating 3D Mesh with Llama 3.1 8Bresearch.nvidia.com · 2024-11-17
  13. ArticleMeta Llama 3.1 405Bopenrouter.ai · 2024-07-23
  14. ArticleCoding with Llama 3.1, New DeepSeek Coder and Mistral Largeaider.chat · 2024-07-25
  15. ArticleLlama-3.1 supports tool calls via promptingbraintrust.dev · 2024-07-26
  16. ArticleMeta releases an open-weights GPT-4-level AI model, Llama 3.1 405Barstechnica.com · 2024-07-23
  17. ArticleShow HN: Run Llama 3.1 8B in the browserapp.wiz.chat · 2024-07-28
  18. VideoRunning the latest Llama 3.1 8B on Raspberry Pi [video]youtube.com · 2024-07-30
  19. ArticleShow HN: Finetune Llama-3.1 2x faster in a Colabcolab.research.google.com · 2024-07-23
  20. ArticleBenchmarking Llama 3.1 405B on 8x AMD MI300X GPUsdstack.ai · 2024-10-09
  21. DiscussionReflection-Llama-3.1-70B is Llama-3 with LoRAold.reddit.com · 2024-09-08
  22. ArticleNvidia's new Llama-3.1 Nemotron Ultra outperforms DeepSeek R1 at half the sizeventurebeat.com · 2025-04-11
  23. ArticleChat with Meta Llama 3.1 405Bllama3.replicate.dev · 2024-07-23

Where it's used

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: 37 quantized builds and 100 Spaces.

  • As a service0no API provider listed
  • On own hardware37quantized builds
  • Built on4225 fine-tunes · 17 adapters · 0 merges
  • In apps100+Hugging Face Spaces

Top Spaces running it: huggingface/paper-central ♥ 230 · aimevzulari/Prompt_Uzmani ♥ 46 · NiansuhAI/HFLLMs ♥ 30 · algorithmicsuperintelligence/OptiLLM ♥ 27

Will it fit?

Weights only, an estimate: leave room for the KV cache. Method

Needs an 8×80 GB node even at INT4

About 244 GB at INT4, 487 GB at FP8, 974 GB at BF16: weights only, so leave room for the KV cache.

16 GB card64 GB Mac2×80 GB8×80 GBINT4: about 244 GB of weights (estimate; KV cache and activations extra)INT4244 GBFP8: about 487 GB of weights (estimate; KV cache and activations extra)FP8487 GBBF16: about 974 GB of weights (estimate; KV cache and activations extra)BF16974 GB
16 GB card64 GB Mac2×80 GB8×80 GBINT4: about 244 GB of weights (estimate; KV cache and activations extra)INT4244 GBFP8: about 487 GB of weights (estimate; KV cache and activations extra)FP8487 GBBF16: about 974 GB of weights (estimate; KV cache and activations extra)BF16974 GB

Estimate params × bytes per param × 1.2 · how we estimate

The story so far

Weights, every dated article, and repo commits on one rail.

10 dated articles and 0 repo commits so far

Teal is the weights (2024-07-16), amber triangles are 10 dated articles (each links to it).

AugSepOctNovDecJanFebMarAprMayJunJulAugSepOctNovDecJanFebMarAprMayJunJulAugSepOctLlama 3.1 · 2024-07-23Llama 3.1 405B now runs at 969 tokens/s on Cerebras Inference · 2024-11-19Meta's Llama 3.1 can recall 42 percent of the first Harry Potter book · 2025-06-15Show HN: Tune LLaMa3.1 on Google Cloud TPUs · 2024-09-11Llama 3.1 in C · 2024-07-24Longwriter – Increase llama3.1 output to 10k words · 2024-10-07Hermes 3: The First Fine-Tuned Llama 3.1 405B Model · 2024-08-15Llama 3.1: Our most capable models to date · 2024-07-23Cerebras launches inference for Llama 3.1; benchmarked at 1846 tokens/s on 8B · 2024-08-27Why Llama 3.1 is Important · 2024-07-23Weights 2024-07-16 · 2024-07-16Weights 2024-07-16now
AugSepOctNovDecJanFebMarAprMayJunJulAugSepOctNovDecJanFebMarAprMayJunJulAugSepOctLlama 3.1 · 2024-07-23Llama 3.1 405B now runs at 969 tokens/s on Cerebras Inference · 2024-11-19Meta's Llama 3.1 can recall 42 percent of the first Harry Potter book · 2025-06-15Show HN: Tune LLaMa3.1 on Google Cloud TPUs · 2024-09-11Llama 3.1 in C · 2024-07-24Longwriter – Increase llama3.1 output to 10k words · 2024-10-07Hermes 3: The First Fine-Tuned Llama 3.1 405B Model · 2024-08-15Llama 3.1: Our most capable models to date · 2024-07-23Cerebras launches inference for Llama 3.1; benchmarked at 1846 tokens/s on 8B · 2024-08-27Why Llama 3.1 is Important · 2024-07-23Weights 2024-07-16 · 2024-07-16Weights 2024-07-16now

What people built on it

79 derivatives on the Hub. Which build to run, by runtime: the most downloaded in each class.

79 derivatives on the Hub, led by 37 quantized

From the Hugging Face model tree; quantized builds in teal, fine-tunes, merges and adapters in greys.

37 quantized (Hugging Face model tree)25 fine-tunes (Hugging Face model tree)17 adapters (Hugging Face model tree)37 quantized25 fine-tunes17 adapters
37 quantized (Hugging Face model tree)25 fine-tunes (Hugging Face model tree)17 adapters (Hugging Face model tree)37 quantized25 fine-tunes17 adapters
GGUF MaziyarPanahi/Meta-Llama-3.1-405B-Instruct-GGUF for llama.cpp, Ollama, LM Studio 7.5M downloads MLX mlx-community/Meta-Llama-3.1-405B-Instruct-8bit for Apple silicon 844k downloads FP8 meta-llama/Llama-3.1-405B-Instruct-FP8 for vLLM, SGLang 554.9k downloads

All variants →

Reference

Everything else, folded: the full tables, every source.

Use-case evidence

1 use case

What it is good at 1 use case

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.

Facts

17 fields

Listed in LLMs

Facts

Pipeline tagtext-generation
Librarytransformers
ArchitecturesLlamaForCausalLM
Model typellama
Params by dtypeBF16: 405,853,388,800
Size bandXXL
MoEno
Gatedmanual
PapersarXiv:2204.05149
LicenseLlama 3.1 Community License (llama3.1)
Hub snapshot2026-10-25

Servability

vLLM · SGLang

Servability

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.

Hardware fit

3 estimates

Hardware fit ESTIMATE

Weights-only memory for Llama-3.1-405B-Instruct: params × bytes/param × 1.2 overhead. KV cache and activations are extra. How this is computed.

BF16 (2 bytes/param)974 GB · Cluster (≥640 GB)
FP8 (1 byte/param)487 GB · Multi-GPU node (<640 GB)
INT4 (0.5 bytes/param)244 GB · Multi-GPU node (<640 GB)

Weights only (params × bytes/param × 1.2 overhead); KV cache and activations extra.

All variants

40 variants

Lineage

Fine-tuned from meta-llama/Llama-3.1-405B (same organization) Family: meta-llama/Llama-3.1 79 derivatives on the Hub

Base model: meta-llama/Llama-3.1-405B

quantized: 37finetune: 25adapter: 17

Top variants (40 shown, ranked by downloads)

Relation
Quant
Size
40 rows
Variant Relation Quant Params Band License Author Likes Downloads
MaziyarPanahi/Meta-Llama-3.1-405B-Instruct-GGUF quantized GGUF — — llama3.1 MaziyarPanahi 15 7.5M
mlx-community/Meta-Llama-3.1-405B-Instruct-8bit quantized -8bit 406B XXL llama3.1 mlx-community 1 844k
meta-llama/Llama-3.1-405B-Instruct-FP8 quantized -FP8 406B XXL llama3.1 meta-llama 193 554.9k
amd/Llama-3.1-405B-Instruct-FP8-KV quantized -FP8 406B XXL other amd 5 129.7k
nvidia/Llama-3.1-405B-Instruct-FP8 finetune -FP8 406B XXL llama3.1 nvidia 19 118.2k
RedHatAI/Meta-Llama-3.1-405B-Instruct-FP8-dynamic quantized -FP8 406B XXL llama3.1 RedHatAI 15 79.6k
nvidia/Llama-3.1-405B-Instruct-NVFP4 finetune NVFP4 230B XXL llama3.1 nvidia 19 71.2k
RedHatAI/Meta-Llama-3.1-405B-Instruct-FP8 quantized -FP8 406B XXL llama3.1 RedHatAI 31 46.4k
amd/Llama-3.1-405B-Instruct-MXFP4-Preview quantized MXFP4 218B XXL llama3.1 amd 1 34.2k
RedHatAI/Meta-Llama-3.1-405B-Instruct-quantized.w4a16 quantized -quant 406B XXL llama3.1 RedHatAI 11 25.1k
nisten/meta-405b-instruct-cpu-optimized-gguf quantized gguf — — — nisten 40 21.5k
hugging-quants/Meta-Llama-3.1-405B-Instruct-BNB-NF4 quantized -BNB 423B XXL llama3.1 hugging-quants 5 6.9k
etemiz/Llama-3.1-405B-Inst-GGUF quantized GGUF — — llama3.1 etemiz 4 6.1k
RedHatAI/Meta-Llama-3.1-405B-Instruct-quantized.w8a8 quantized -quant 406B XXL llama3.1 RedHatAI 2 4.5k
RedHatAI/Meta-Llama-3.1-405B-Instruct-quantized.w8a16 quantized -quant 406B XXL llama3.1 RedHatAI 2 2.4k
shisa-ai/shisa-v2-llama3.1-405b finetune — 406B XXL llama3.1 shisa-ai 19 1.3k
uiuc-convai/CoALM-405B finetune — 406B XXL cc-by-nc-4.0 uiuc-convai 7 1.6k
mlabonne/BigLlama-3.1-681B-Instruct finetune — — — — mlabonne 12 977
mlx-community/Llama-3.1-405B-Instruct-8bit quantized -8bit 406B XXL llama3.1 mlx-community 2 1.2k
jplhughes2/1a_meta-llama-Llama-3.1-405B-Instruct-fsdp-lr1e-5 adapter — — — llama3.1 jplhughes2 1 1.2k
jkim96/Llama-3.1-405B-Instruct-DASHQ-INT2-g32 quantized — 4.21B M llama3.1 jkim96 0 785
jburmeister/Meta-Llama-3.1-405B-Instruct-GGUF quantized GGUF — — llama3.1 jburmeister 0 693
Snowflake/Llama-3.1-SwiftKV-405B-Instruct-FP8 quantized -FP8 425B XXL llama3.1 Snowflake 0 680
nicoboss/Meta-Llama-3.1-405B-Instruct-Uncensored-Lora adapter — — — llama3.1 nicoboss 9 204
ApolloRaines/Llama-3.1-405B-Instruct-Jbliterated finetune — 406B XXL llama3.1 ApolloRaines 1 570
nicoboss/Meta-Llama-3.1-405B-Instruct-Uncensored adapter — 406B XXL llama3.1 nicoboss 10 0
IntuitIntel/Llama-3.1-405B-Instruct-8bit quantized -8bit 406B XXL llama3.1 IntuitIntel 0 485
VPTQ-community/Meta-Llama-3.1-405B-Instruct-v16-k65536-64-woft quantized — 22.4B L llama3.1 VPTQ-community 3 326
xmadai/Llama-3.1-405B-Instruct-xMADai-INT4 quantized INT4 406B XXL llama3.1 xmadai 6 173
VPTQ-community/Meta-Llama-3.1-405B-Instruct-v16-k65536-65536-woft quantized — 31.2B L llama3.1 VPTQ-community 3 306
relaxml/Llama-3.1-405B-Instruct-QTIP-2Bit finetune — 54.5B XL — relaxml 3 184
mradermacher/Meta-Llama-3.1-405B-Instruct-GGUF finetune GGUF — — llama3.1 mradermacher 6 0
VPTQ-community/Meta-Llama-3.1-405B-Instruct-v16-k32768-32768-woft quantized — 28.7B L llama3.1 VPTQ-community 1 246
VPTQ-community/Meta-Llama-3.1-405B-Instruct-v16-k65536-256-woft quantized — 24.0B L llama3.1 VPTQ-community 1 244
VPTQ-community/Meta-Llama-3.1-405B-Instruct-v16-k65536-128-woft quantized — 23.2B L llama3.1 VPTQ-community 1 239
VPTQ-community/Meta-Llama-3.1-405B-Instruct-v16-k65536-1024-woft quantized — 25.6B L llama3.1 VPTQ-community 1 235
VPTQ-community/Meta-Llama-3.1-405B-Instruct-v8-k4096-0-woft quantized — 23.1B L llama3.1 VPTQ-community 1 230
OPEA/Meta-Llama-3.1-405B-Instruct-int4-sym-inc quantized int4 55.0B XL llama3.1 OPEA 1 211
VPTQ-community/Meta-Llama-3.1-405B-Instruct-v8-k65536-256-woft quantized — 42.4B XL llama3.1 VPTQ-community 1 179
VPTQ-community/Meta-Llama-3.1-405B-Instruct-v8-k65536-65536-woft quantized — 55.4B XL llama3.1 VPTQ-community 0 225

Articles and coverage

25 pieces

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.

Official3

Llama 3.1
llama.meta.com2024-07-23officialHN front pageabout the Llama 3.1 familyHN 437 pts · 269 comments42 words
Llama 3.1 405B now runs at 969 tokens/s on Cerebras Inference
cerebras.ai2024-11-19HN front pageHN 427 pts · 156 comments553 words
Cerebras Inference runs Llama 3.1 405B at 969 tokens per second, achieving the highest recorded performance at 128K context length. The platform delivers a 240ms time-to-first-token and offers general availability in Q1 2025 at $6 per million input tokens and $12 per million output tokens. This result is 12x faster than the best GPU result and 75x faster than AWS.written, reviewed, cited
Llama 3.1: Our most capable models to date
ai.meta.com2024-07-23officialauthorityabout the Llama 3.1 familyHN 27 pts · 3 comments2,136 words

Analysis6

Meta's Llama 3.1 can recall 42 percent of the first Harry Potter book
understandingai.org2025-06-15HN front pageabout the Llama 3.1 familyHN 191 pts · 305 comments
Show HN: Tune LLaMa3.1 on Google Cloud TPUs
github.com2024-09-11authorityHN front pageabout the Llama 3.1 familyHN 189 pts · 52 comments869 words
Llama 3.1 in C
github.com2024-07-24authorityHN front pageabout the Llama 3.1 familyHN 212 pts · 36 comments5,858 words
Longwriter – Increase llama3.1 output to 10k words
github.com2024-10-07authorityHN front pageabout the Llama 3.1 familyHN 154 pts · 29 comments6,696 words
Hermes 3: The First Fine-Tuned Llama 3.1 405B Model
lambdalabs.com2024-08-15HN front pageHN 146 pts · 67 comments530 words
Nous Research launched Hermes 3, the first full-parameter fine-tune of Meta's Llama 3.1 405B model, trained on an 8-node Lambda 1-Click Cluster. The model achieves a 50% reduction in VRAM and disk requirements through Neural Magic's FP8 quantization, enabling single-node operation. Hermes 3, which has been downloaded over 33 million times, is available for free via Lambda's Chat Completions API and Lambda Chat interface.written, reviewed, cited
Why Llama 3.1 is Important
timkellogg.me2024-07-23about the Llama 3.1 familyHN 56 pts · 23 comments

Community1

Cerebras launches inference for Llama 3.1; benchmarked at 1846 tokens/s on 8B
twitter.com2024-08-27about the Llama 3.1 familyHN 95 pts · 42 comments
All coverage tracked for this model (25) — append-only, nothing is ever removed
TitleSourcePublishedFirst seenSignals
Show HN: Llama 3.1 70B on a single RTX 3090 via NVMe-to-GPU bypassing the CPUgithub.com2026-02-212026-10-07hn
Meta's Llama 3.1 can recall 42 percent of the first Harry Potter bookunderstandingai.org2025-06-152026-10-07hn
Nvidia's new Llama-3.1 Nemotron Ultra outperforms DeepSeek R1 at half the sizeventurebeat.com2025-04-112026-10-07hn
Llama 3.1 405B now runs at 969 tokens/s on Cerebras Inferencecerebras.ai2024-11-192026-10-07hn
Nvidia presents Llama-Mesh: Generating 3D Mesh with Llama 3.1 8Bresearch.nvidia.com2024-11-172026-10-07hn
Benchmarking Llama 3.1 405B on 8x AMD MI300X GPUsdstack.ai2024-10-092026-10-07hn
Longwriter – Increase llama3.1 output to 10k wordsgithub.com2024-10-072026-10-07hn
Llama 3.1 Omni Modelgithub.com2024-09-182026-10-07hn
g1: Using Llama-3.1 70B on Groq to create o1-like reasoning chainsgithub.com2024-09-152026-10-07hn
Show HN: Tune LLaMa3.1 on Google Cloud TPUsgithub.com2024-09-112026-10-07hn
Reflection-Llama-3.1-70B is Llama-3 with LoRAold.reddit.com2024-09-082026-10-07hn
Cerebras launches inference for Llama 3.1; benchmarked at 1846 tokens/s on 8Btwitter.com2024-08-272026-10-07hn
Hermes 3: The First Fine-Tuned Llama 3.1 405B Modellambdalabs.com2024-08-152026-10-07hn
Running the latest Llama 3.1 8B on Raspberry Pi [video]youtube.com2024-07-302026-10-07hn
Show HN: Run Llama 3.1 8B in the browserapp.wiz.chat2024-07-282026-10-07hn
Llama-3.1 supports tool calls via promptingbraintrust.dev2024-07-262026-10-07hn
Coding with Llama 3.1, New DeepSeek Coder and Mistral Largeaider.chat2024-07-252026-10-07hn
Llama 3.1 in Cgithub.com2024-07-242026-10-07hn
Why Llama 3.1 is Importanttimkellogg.me2024-07-232026-10-07hn
Show HN: Finetune Llama-3.1 2x faster in a Colabcolab.research.google.com2024-07-232026-10-07hn
Meta releases an open-weights GPT-4-level AI model, Llama 3.1 405Barstechnica.com2024-07-232026-10-07hn
Meta Llama 3.1 405Bopenrouter.ai2024-07-232026-10-07hn
Chat with Meta Llama 3.1 405Bllama3.replicate.dev2024-07-232026-10-07hn
Llama 3.1: Our most capable models to dateai.meta.com official2024-07-232026-10-07hn
Llama 3.1llama.meta.com official2024-07-232026-10-07hn

Papers

1 paper

Papers 1

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).

Similar models

3 models

Similar models

  • Hermes-3-Llama-3.1-405B · 406B
    same size band (XXL), same task (text-generation), more providers
  • GLM-4.7 · 358B
    same size band (XXL), same task (text-generation), more providers, permissive license
  • GLM-4.6 · 357B
    same size band (XXL), same task (text-generation), more providers, permissive license

Sources

3 sources

Sources

Hub snapshot: 2026-10-25 · HF API: 2026-10-24 21:38 UTC · Providers: 2026-10-24 21:44 UTC · Articles: 2026-10-24 21:46 UTC · Summary: 2026-10-07 15:08 UTC · Last verified 2026-10-25

Follow Llama-3.1-405B-Instruct

Watch this model

Email alerts are not open yet; the feeds above are the same data. See feeds for the item schema.