The story so far
Weights, every dated article, and repo commits on one rail.
30 repo commits since the weights landed
Teal is the weights (2024-04-18), grey ticks are repo commits.
saiga_llama3_8b
IlyaGusev Llama Community License (llama3) out 3 years ago
8.03B parameters, text in, text out, 8K tokens of context. 1 provider serves it.
Verified 2026-10-25 Rank #751 in LLMs · score 71.5 method v1 Sources 3 Downloads 30d 396k · #183 of 619 tracked
Do people use it?396kdownloads, 30 days#183 of 619 we track · top 30%+0 likes, downloads −2% this week144 likesHugging Face2026-10-24
Where to run it
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 |
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.
Served by 1 API provider; 28 quantized builds for local use.
Will it fit?
Weights only, an estimate: leave room for the KV cache. Method
Fits a 16 GB card at INT4
About 4.8 GB at INT4, 9.6 GB at FP8, 19 GB at BF16: weights only, so leave room for the KV cache.
Estimate params × bytes per param × 1.2 · how we estimate
What people built on it
58 derivatives on the Hub. Which build to run, by runtime: the most downloaded in each class.
58 derivatives on the Hub, led by 28 quantized
From the Hugging Face model tree; quantized builds in teal, fine-tunes, merges and adapters in greys.
Reference
Everything else, folded: the full tables, every source.
Benchmarks
8 rows
Benchmarks 8 rows
measured third-party leaderboard rowauthor-reported number from the model card or eval files (8)No confidence intervals unless the source gave them; settings differ between rows, so compare within a benchmark, not across. Date = the benchmark's own date when it gives one, else when we read it.
| Benchmark | Value | Setting / notes | Reported by | Date |
|---|---|---|---|---|
| author-reported AlpacaEval · Chat and follow instructions | 68.31 percent | Pivot: chatgpt_3_5_turbo, saiga_llama3_8b, v6 | model author | 2026-10-07 |
| author-reported AlpacaEval · Chat and follow instructions | 69.4 percent | Pivot: chatgpt_3_5_turbo, saiga_llama3_8b, v7 | model author | 2026-10-07 |
| author-reported AlpacaEval · Chat and follow instructions | 66.18 percent | Pivot: chatgpt_3_5_turbo, saiga_llama3_8b, v5 | model author | 2026-10-07 |
| author-reported AlpacaEval · Chat and follow instructions | 65.9 percent | Pivot: chatgpt_3_5_turbo, saiga_llama3_8b, v4 | model author | 2026-10-07 |
| author-reported AlpacaEval · Chat and follow instructions | 61.08 percent | Pivot: chatgpt_3_5_turbo, saiga_llama3_8b, v3 | model author | 2026-10-07 |
| author-reported AlpacaEval · Chat and follow instructions | 48.19 percent | Pivot: chatgpt_3_5_turbo, saiga_llama3_8b, v2 | model author | 2026-10-07 |
| author-reported AlpacaEval · Chat and follow instructions | 49.16 percent | Pivot: sfr, saiga_llama3_8b, v7 | model author | 2026-10-07 |
| author-reported AlpacaEval · Chat and follow instructions | 47.23 percent | Pivot: sfr, saiga_llama3_8b, v6 | model author | 2026-10-07 |
What the card says
1 cited sentence
Use-case evidence
2 use cases
What it is good at 2 use cases
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
20 fields
Listed in LLMs
Facts
- saiga_llama3_8b has 8.03B parameters. (safetensors metadata, observed 2026-10-24)
- Context length is 8,192 tokens (8K). (config.json max_position_embeddings, observed 2026-10-24)
- License: Llama Community License (llama3) (friction: attribution). (license file, observed 2026-10-24)
- Created 2024-04-18, last modified 2024-07-04. (HF API, observed 2026-10-24)
| Pipeline tag | text-generation |
|---|---|
| Library | transformers |
| Architectures | LlamaForCausalLM |
| Model type | llama |
| Dtype | bfloat16 |
| Params by dtype | BF16: 8,030,261,248 |
| Size band | M |
| MoE | no |
| Context length | 8,192 (config.json max_position_embeddings) |
| Languages | ru |
| License | Llama Community License (llama3) (other) |
| Hub snapshot | 2026-10-25 |
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 saiga_llama3_8b: params × bytes/param × 1.2 overhead. KV cache and activations are extra. How this is computed.
| BF16 (2 bytes/param) | 19 GB · Single GPU (<24 GB) |
|---|---|
| FP8 (1 byte/param) | 9.6 GB · Single GPU (<24 GB) |
| INT4 (0.5 bytes/param) | 4.8 GB · Single GPU (<24 GB) |
Weights only (params × bytes/param × 1.2 overhead); KV cache and activations extra.
All variants
40 variants
Lineage
quantized: 28finetune: 9merge: 10adapter: 11
Top variants (40 shown, ranked by downloads)
| Variant | Relation | Quant | Params | Band | License | Author | Likes | Downloads |
|---|---|---|---|---|---|---|---|---|
| QuantFactory/saiga_llama3_8b-GGUF | quantized | GGUF | — | — | other | QuantFactory | 7 | 25.3k |
| deepvk/llava-saiga-8b | finetune | — | 8.36B | M | apache-2.0 | deepvk | 16 | 6.8k |
| itlwas/saiga_llama3_8b-Q4_K_M-GGUF | quantized | GGUF | — | — | other | itlwas | 0 | 5.1k |
| mradermacher/saiga_llama3_8b-GGUF | quantized | GGUF | — | — | other | mradermacher | 2 | 4.8k |
| Content-AI/saiga_llama3_8b-Q4_K_M-GGUF | quantized | GGUF | — | — | other | Content-AI | 0 | 3.6k |
| ehristoforu/llama-3-12b-instruct | merge | — | 12.0B | L | apache-2.0 | ehristoforu | 6 | 3.1k |
| tensorblock/saiga_llama3_8b-GGUF | quantized | GGUF | — | — | other | tensorblock | 1 | 3.3k |
| d0rj/Llama-3-8B-saiga-suzume-ties | merge | — | 8.03B | M | llama3 | d0rj | 6 | 3.1k |
| featherless-ai-quants/IlyaGusev-saiga_llama3_8b-GGUF | quantized | GGUF | — | — | — | featherless-ai-quants | 0 | 2.9k |
| QuantFactory/Multilingual-SaigaSuzume-8B-GGUF | merge | GGUF | — | — | — | QuantFactory | 2 | 2.7k |
| liodon-ai/saiga_llama3_8b-imatrix-GGUF | quantized | imatrix | — | — | other | liodon-ai | 0 | 2.2k |
| RachidAR/Llama-3-8B-saiga-suzume-ties-Q6_K-GGUF-OLD | merge | Q6_K | — | — | llama3 | RachidAR | 2 | 1.9k |
| RachidAR/saiga_llama3_8b-Q6_K-GGUF | quantized | Q6_K | — | — | other | RachidAR | 0 | 1.8k |
| Futyn-Maker/saiga_llama3_8b_wildberries_4bit_gguf | quantized | gguf | — | — | apache-2.0 | Futyn-Maker | 2 | 1.3k |
| qucklecrabik/model | quantized | GGUF | — | — | apache-2.0 | qucklecrabik | 0 | 992 |
| VlSav/saiga_llama3_8b_v7-Q6_K-GGUF | quantized | Q6_K | — | — | other | VlSav | 1 | 831 |
| evgensoft/saiga_llama3_8b-Q5_K_M-GGUF | quantized | GGUF | — | — | other | evgensoft | 1 | 700 |
| sdfjgdskfds/saiga_llama3_8b-Q4_K_M-GGUF | quantized | GGUF | — | — | other | sdfjgdskfds | 0 | 734 |
| Khetterman/CursedMatrix-8B-v9 | merge | — | 8.03B | M | — | Khetterman | 6 | 422 |
| compressa-ai/Saiga-Llama-3-8B-OmniQuant | quantized | — | 8.03B | M | other | compressa-ai | 0 | 720 |
| Khetterman/Kosmos-8B-v1 | merge | — | 8.03B | M | — | Khetterman | 4 | 402 |
| Khetterman/Multilingual-SaigaSuzume-8B | merge | — | 8.03B | M | — | Khetterman | 2 | 333 |
| metroproliv/saiga_for_TW | quantized | GGUF | — | — | apache-2.0 | metroproliv | 0 | 362 |
| emile778/saiga_llama3_8b-Q4_0-GGUF | quantized | Q4_0 | — | — | other | emile778 | 0 | 330 |
| vladlen32230/Saiga-llama3-L3-Stheno-8B-v3.2 | merge | — | 8.03B | M | mit | vladlen32230 | 1 | 200 |
| steamdroid/saiga_llama3_8b-mlx-4Bit | quantized | mlx | 8.03B | M | other | steamdroid | 0 | 216 |
| foxxar04/saiga-llama3-8b-4bit | quantized | -4bit | 8.25B | M | llama3.1 | foxxar04 | 0 | 161 |
| ch1eph/geosaiga_llama3_8b_gguf | quantized | gguf | — | — | — | ch1eph | 0 | 148 |
| alexgusevski/saiga_llama3_8b-q4-mlx | quantized | mlx | 8.03B | M | other | alexgusevski | 0 | 125 |
| alexgusevski/saiga_llama3_8b-mlx | finetune | mlx | 8.03B | M | other | alexgusevski | 0 | 103 |
| alexgusevski/saiga_llama3_8b-q8-mlx | quantized | mlx | 8.03B | M | other | alexgusevski | 0 | 100 |
| memstre24/invision-weights-lora | adapter | — | — | — | — | memstre24 | 0 | 93 |
| liodon-ai/saiga_llama3_8b-ONNX | quantized | — | — | — | other | liodon-ai | 0 | 89 |
| alexgusevski/saiga_llama3_8b-q2-mlx | quantized | mlx | 8.03B | M | other | alexgusevski | 0 | 83 |
| alexgusevski/saiga_llama3_8b-q6-mlx | quantized | mlx | 8.03B | M | other | alexgusevski | 0 | 80 |
| alexgusevski/saiga_llama3_8b-q3-mlx | quantized | mlx | 8.03B | M | other | alexgusevski | 0 | 73 |
| ehristoforu/0001lp | merge | — | 12.0B | L | — | ehristoforu | 0 | 72 |
| paramedik/saiga_llama3_8b-openvino | finetune | — | — | — | other | paramedik | 0 | 65 |
| cringoulique/maxprimodel | finetune | — | — | — | apache-2.0 | cringoulique | 0 | 64 |
| artemgoncarov/saiga_llama_8b_tinkoff_2 | quantized | GGUF | — | — | apache-2.0 | artemgoncarov | 0 | 62 |
Repository changelog
30 commits
Repository changelog 30 commits
| Date | Commit | SHA |
|---|---|---|
| 2024-07-04 | Update README.md | 5bb9917 |
| 2024-07-04 | Update README.md | 7fa5874 |
| 2024-07-04 | Update README.md | 890b148 |
| 2024-07-03 | Update README.md | 3741434 |
| 2024-07-03 | Update README.md | 1c192fc |
| 2024-07-02 | v7 aka saiga_llama3_8b_abliterated_sft_m11_d7_kto_m7_d7 | 17e1bcc |
| 2024-06-01 | Update README.md | 476f1d5 |
| 2024-06-01 | Update README.md | 6ffd860 |
| 2024-06-01 | v6 | b662833 |
| 2024-05-28 | Update README.md | bc2b4f9 |
| 2024-05-28 | Update README.md | 46a2dac |
| 2024-05-22 | Update README.md | 256d917 |
| 2024-05-22 | Update README.md | 3f7de9b |
| 2024-05-22 | Update README.md | 06693a4 |
| 2024-05-22 | Update README.md | 46a4d31 |
| 2024-05-22 | Update README.md | 636d399 |
| 2024-05-22 | v5 | d947b00 |
| 2024-04-26 | Update README.md | 6615aef |
| 2024-04-26 | Update README.md | 1cc945d |
| 2024-04-26 | Update README.md | 3fed7fc |
| 2024-04-26 | v4: ChatML -> LLama3 template | bc54c81 |
| 2024-04-25 | Update README.md | c588356 |
| 2024-04-25 | Tagengo version | 91d7b8a |
| 2024-04-20 | Update README.md | ae61b4f |
| 2024-04-19 | Update README.md | bc94b1d |
| 2024-04-19 | Update chat template | d04705b |
| 2024-04-19 | Update README.md | 55a2da9 |
| 2024-04-19 | Better generation config | 0af818c |
| 2024-04-18 | Update README.md | 07d82ec |
| 2024-04-18 | Update README.md | 1e84ce0 |
All providers
1 provider · 1 endpoint
Where to run 1 providers · 1 endpoints
Hosted endpoints observed for saiga_llama3_8b, 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 | IlyaGusev/saiga_llama3_8b | — | — | — | — | — | — | — | live | 2026-10-25 |
Similar models
3 models
Similar models
- Llama-3.1-8B-Instruct · 8.03B same size band (M), same task (text-generation), more providers
- Qwen3-8B · 8.19B same size band (M), same task (text-generation), more providers, permissive license
- Meta-Llama-3-8B-Instruct · 8.03B same size band (M), same task (text-generation), more providers
Sources
3 sources
Sources
Hub snapshot: 2026-10-25 · HF API: 2026-10-24 21:43 UTC · Providers: 2026-10-24 21:44 UTC · Articles: 2026-10-24 21:46 UTC · Summary: 2026-10-07 15:15 UTC · Last verified 2026-10-25
Follow saiga_llama3_8b
Watch this model
- /feeds/models/IlyaGusev--saiga_llama3_8b.xml(RSS 2.0) · every change to saiga_llama3_8b
- /feeds/families/IlyaGusev--saiga.xml(RSS 2.0) · every change across the saiga family
Email alerts are not open yet; the feeds above are the same data. See feeds for the item schema.