Up to date
NousResearch Llama 3 Community License out 2 years ago
406B parameters, text in, text out, 128K tokens of context. 1 provider serves it from $1.00 per million input tokens.
Verified 2026-10-25 Rank #619 in LLMs · score 73.7 method v1 Sources 4 Downloads 30d 1.4k · #587 of 619 tracked
Picked by rule from 4 tracked pieces: the lab's own post, the most shared, a field report, the latest.
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 2 picks above, plus these 2. Top is the radar's score: discussion, source authority and the lab's own posts, fading with age.
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 | DeepInfra | $1.00 | $1.00 | 128K | fp8 | Oct 24 |
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; 3 quantized builds for local use.
Top Spaces running it: Nymbo/overthink-1 ♥ 35 · algorithmicsuperintelligence/OptiLLM ♥ 27 · wenbopan/overthink-1 ♥ 11 · patched-codes/analyze-gh-repo ♥ 2
Weights only, an estimate: leave room for the KV cache. Method
About 244 GB at INT4, 487 GB at FP8, 974 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 (2024-08-13), amber triangles are 4 dated articles (each links to it), grey ticks are repo commits.
11 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.
Hermes 3 is a 406B parameter text-generation finetune of the Llama-3.1 405B foundation model.[1][2] Developed by NousResearch, this model uses the ChatML prompt format and is released under the Llama 3 Community License.[3][4][5]
Generated 2026-10-07 15:14 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 | LlamaForCausalLM |
| Model type | llama |
| Dtype | bfloat16 |
| Params by dtype | BF16: 405,853,388,800 |
| Size band | XXL |
| MoE | no |
| Context length | 131,072 (config.json max_position_embeddings) |
| Languages | en |
| Papers | arXiv:2408.11857 |
| License | Llama 3 Community License (llama3) |
| 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 Hermes-3-Llama-3.1-405B: 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.
Base model: meta-llama/Llama-3.1-405B
quantized: 3adapter: 2finetune: 5merge: 1
| Variant | Relation | Quant | Params | Band | License | Author | Likes | Downloads |
|---|---|---|---|---|---|---|---|---|
| bartowski/Hermes-3-Llama-3.1-405B-GGUF | quantized | GGUF | — | — | llama3 | bartowski | 14 | 47.4k |
| DevQuasar/NousResearch.Hermes-3-Llama-3.1-405B-GGUF | quantized | GGUF | — | — | — | DevQuasar | 3 | 16.5k |
| nicoboss/Hermes-3-Llama-3.1-405B-Uncensored | adapter | — | 406B | XXL | llama3.1 | nicoboss | 46 | 0 |
| mrtoots/Hermes-3-Llama-3.1-405B-mlx-2Bit | quantized | mlx | 406B | XXL | llama3 | mrtoots | 0 | 744 |
| mradermacher/Hermes-3-Llama-3.1-405B-GGUF | finetune | GGUF | — | — | llama3 | mradermacher | 2 | 0 |
| mradermacher/Hermes-3-Llama-3.1-405B-i1-GGUF | finetune | GGUF | — | — | llama3 | mradermacher | 2 | 0 |
| Abigail45/Forest | merge | — | — | — | apache-2.0 | Abigail45 | 1 | 0 |
| SipsaLabs/hermes-3-llama-3.1-405b-uc-v3-bpw5 | finetune | — | — | — | other | SipsaLabs | 1 | 0 |
| ruinedbythorns/SabrinaQV2 | finetune | — | — | — | mit | ruinedbythorns | 1 | 0 |
| nicoboss/Hermes-3-Llama-3.1-405B-Uncensored-Lora | adapter | — | — | — | llama3.1 | nicoboss | 0 | 15 |
| SipsaLabs/hermes-3-llama-3.1-405b-streaming-bpw5 | finetune | — | — | — | other | SipsaLabs | 0 | 0 |
| Date | Commit | SHA |
|---|---|---|
| 2024-10-08 | Add proper library name (#3) | b9fe8ed |
| 2024-09-08 | fixing tool_use jinja template bugs | e856e74 |
| 2024-08-23 | Update README.md | 47f8572 |
| 2024-08-23 | Update README.md | d3ce169 |
| 2024-08-22 | Update tokenizer_config.json | 0df25cb |
| 2024-08-18 | Update tokenizer_config.json | 69f35ae |
| 2024-08-15 | Update README.md | f567b06 |
| 2024-08-15 | Update README.md | e279392 |
| 2024-08-15 | Update README.md | 5017d81 |
| 2024-08-15 | Update README.md | 831b09a |
| 2024-08-15 | Update README.md | 2c4c029 |
| 2024-08-15 | Update README.md | 503828b |
| 2024-08-15 | Update README.md | e67e35c |
| 2024-08-14 | Update README.md | 827c6a0 |
| 2024-08-14 | Create README.md | a044473 |
| 2024-08-14 | Update tokenizer_config.json | 500ce36 |
| 2024-08-14 | Update config.json | 2493e45 |
| 2024-08-14 | Update generation_config.json | d76be33 |
| 2024-08-13 | Upload folder using huggingface_hub | 96f8d55 |
| 2024-08-13 | Update generation_config.json | 660d6f9 |
| 2024-08-13 | Update config.json | ff42a2f |
| 2024-08-13 | Upload folder using huggingface_hub | 8778f05 |
| 2024-08-13 | initial commit | 08131d0 |
Hosted endpoints observed for Hermes-3-Llama-3.1-405B, sorted by input price (USD per million tokens). Each row names its observer and date; prices change daily, so check the source before quoting. Cheapest observed input price: $1.00/M.
| Provider | Model ref | Context | Quant | $/M in | $/M out | tok/s | TTFT | Tools | Status | Observed |
|---|---|---|---|---|---|---|---|---|---|---|
| DeepInfra openrouter | nousresearch/hermes-3-llama-3.1-405b | 128K | fp8 | $1.00 | $1.00 | — | — | — | live | 2026-10-24 |
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 21:42 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.
