Llama 3.1 8B
A 8 billion-parameterHow many numbers the model learned during training — the usual rough measure of its size. dense model from Meta. 8B, 70B and 405B models with 128K context.
- Size
- 8.03Bdense
- Memory to run
- ~6 GB estimatesmallest, 8K ctx
- Context
- 128Ktokens
- License
- PermissiveLlama 3.1 Community License; Apache License 2.0
- Updated
- 18 Sept 2026
Where it runs
Best-fitting version on each reference system, at an 8K contextHow much text the model can consider at once, counted in tokens — roughly ¾ of a word each..
Runs well 16
- A100 80GB server~146 tok/s estimate
- Dual RTX 3090~64 tok/s estimate
- GB10 mini workstation~20 tok/s estimate
- MacBook Air M2 16 GB~7 tok/s estimate
- MacBook Air M3 16 GB~7 tok/s estimate
- MacBook Air M4 16 GB~9 tok/s estimate
- MacBook Pro M3 Max 64 GB54 tok/s measured
- MacBook Pro M4 Max 128 GB~39 tok/s estimate
- Mac mini M4 16 GB~9 tok/s estimate
- Mac mini M4 Pro 48 GB~20 tok/s estimate
- Mac Studio M2 Ultra 192 GB~57 tok/s estimate
- RTX 4060 Ti 16GB budget build~21 tok/s estimate
- RTX 4090 workstation128 tok/s measured
- RTX 5090 workstation~128 tok/s estimate
- RX 7900 XTX desktop94 tok/s measured
- Ryzen AI Max+ 395 mini PC~18 tok/s estimate
Slowly (CPU or offload) 2
- CPU-only Ryzen 9 7950X12.4 tok/s measured
- Windows laptop 16 GB~6 tok/s estimate
Too large 1
Measured throughput
| Quantization | System | Runtime | Context | Measurements | Source |
|---|---|---|---|---|---|
| Llama 3.1 8B Instruct Q4_K_M (LM Studio Community) | CPU-only Ryzen 9 7950X (128 GB DDR5) | llama.cppcpu · b4600 | 4K | 92 tok/s · Prompt processing (512) 12.4 tok/s · Generation (128) | Mutinai illustrative fixtures |
| Llama 3.1 8B Instruct Q4_K_M (LM Studio Community) | MacBook Pro M3 Max 64 GB | llama.cppmetal · b4600 | 4K | 760 tok/s · Prompt processing (512) 54 tok/s · Generation (128) | Mutinai illustrative fixtures |
| Llama 3.1 8B Instruct Q4_K_M (LM Studio Community) | RTX 4090 workstation (64 GB DDR5) | llama.cppcuda · b4600 | 4K | 12,100 tok/s · Prompt processing (512) 128 tok/s · Generation (128) | Mutinai illustrative fixtures |
| Llama 3.1 8B Instruct Q4_K_M (LM Studio Community) | RX 7900 XTX desktop (64 GB DDR5) | llama.cpprocm · b4600 | 4K | 3,050 tok/s · Prompt processing (512) 94 tok/s · Generation (128) | Mutinai illustrative fixtures |
Versions & downloads
Each version is a separate set of weights. Expand one for its downloads: quantizationStoring each of the model’s numbers with fewer bits, so the file is smaller and needs less memory. trades a little quality for a much smaller file, and the memory column adds the working memoryExtra space the model needs while it answers. It grows with the length of the conversation, on top of the file itself. a conversation needs on top.
Variant Llama 3.1 8BRestricted
- Good for
- Publisher
- Meta
- License
- Llama 3.1 Community License · commercial use restricted
- Released
- 23 Jul 2024
- huggingface
- meta-llama/Llama-3.1-8B
Contributions are not open yet, so there is nothing here from members.
| Quantization | Format | Bits | Download | Memory @ 8K | Publisher |
|---|---|---|---|---|---|
| Download BF16native | safetensors | 16 | 15.0 GiB | Metameta-llama/Llama-3.1-8B |
Variant Llama 3.1 8B InstructRestricted
- Good for
- Publisher
- Meta
- License
- Llama 3.1 Community License · commercial use restricted
- Released
- 23 Jul 2024
- huggingface
- meta-llama/Llama-3.1-8B-Instruct
Contributions are not open yet, so there is nothing here from members.
| Quantization | Format | Bits | Download | Memory @ 8K | Publisher |
|---|---|---|---|---|---|
| Download BF16native | safetensors | 16 | 15.0 GiB | Metameta-llama/Llama-3.1-8B-Instruct | |
| Download Q8_0legacy gguf | gguf | 8.5 | 7.9 GiB | LM Studio Communitylmstudio-community/Llama-3.1-8B-Instruct-GGUF | |
| Download FP8fp8 | safetensors | 8.1 | 8.5 GiB | RedHatAIRedHatAI/Meta-Llama-3.1-8B-Instruct-FP8 | |
| Download Q4_K_Mk quant | gguf | 4.89 | 4.6 GiB | LM Studio Communitylmstudio-community/Llama-3.1-8B-Instruct-GGUF | |
| Download MLX 4-bitmlx | mlx | 4.5 | 4.2 GiB | MLX Communitymlx-community/Llama-3.1-8B-Instruct-4bit |
Variant HermesLicense unknown
- Good for
- Publisher
- dedsecisback2026
- License
- Unknown
- Released
- 21 Aug 2026
- huggingface
- dedsecisback2026/Hermes
Contributions are not open yet, so there is nothing here from members.
No downloadable artifacts recorded.
Variant Hermes 3 Llama 3.1 8BRestricted
- Good for
- Publisher
- NousResearch
- License
- Llama 3.1 Community License · commercial use restricted
- Released
- 28 Jul 2024
- huggingface
- NousResearch/Hermes-3-Llama-3.1-8B
Contributions are not open yet, so there is nothing here from members.
| Quantization | Format | Bits | Download | Memory @ 8K | Publisher |
|---|---|---|---|---|---|
| Download Q8_0legacy gguf | gguf | 8.5 | 8.0 GiB | NousResearchNousResearch/Hermes-3-Llama-3.1-8B-GGUF | |
| Download Q6_Kk quant | gguf | 6.56 | 6.1 GiB | NousResearchNousResearch/Hermes-3-Llama-3.1-8B-GGUF | |
| Download Q4_K_Mk quant | gguf | 4.89 | 4.6 GiB | NousResearchNousResearch/Hermes-3-Llama-3.1-8B-GGUF |
Variant neurologist 7B instructPermissive
- Good for
- Publisher
- mainbrains
- License
- Apache License 2.0 · commercial use allowed
- Released
- 20 Aug 2026
- huggingface
- mainbrains/neurologist-7b-instruct
Contributions are not open yet, so there is nothing here from members.
No downloadable artifacts recorded.
Variant solana nvidia trading factory 8BPermissive
- Good for
- Publisher
- solanaclawd
- License
- Apache License 2.0 · commercial use allowed
- Released
- 6 Sept 2026
- huggingface
- solanaclawd/solana-nvidia-trading-factory-8b
Contributions are not open yet, so there is nothing here from members.
No downloadable artifacts recorded.
Other sizes in Llama 3.1: Llama 3.1 70B
Benchmarks
Each benchmarkA fixed set of questions every model is given, so their scores can be compared on the same task. score is labelled with who produced it. Developer-reported scores use each lab’s own prompts and settings; a benchmark that runs every model itself uses one setup for all. Bars are relative to the best open result.
Community
Benchmark runs and reviews from people running it on their own hardware.
Benchmark runs
Reviews
Details & sources
Architecture
- Parameters
- 8.03B
- Active / token
- All (dense)
- Architecture
- Dense
- Layers
- 32
- Attention heads
- 32
- KV heads
- 8
- Head dim
- 128
- KV cache @ 8K (fp16)
- 1.00 GiB
- Max context
- 131,072 tokens
Lineage
How this model’s versions relate to each other and to other models.
Sources
- Hugging Face (fixture) (illustrative fixture, 1 record, 13 Sept 2026)
- Hugging Face Hub (live source, 1 record, 13 Sept 2026)
- Mutinai illustrative fixtures (illustrative fixture, 1 record, 13 Sept 2026)
External identifiers
- huggingface meta-llama/Llama-3.1-8B
- huggingface meta-llama/Llama-3.1-8B-Instruct
- huggingface dedsecisback2026/Hermes
- huggingface NousResearch/Hermes-3-Llama-3.1-8B
- huggingface mainbrains/neurologist-7b-instruct
- huggingface solanaclawd/solana-nvidia-trading-factory-8b
Field history
- license = "llama3.1" — Hugging Face Hub (current)
- license = "llama3.1" — Hugging Face (fixture)
- base = "NousResearch/Hermes-3-Llama-3.1-8B" — Hugging Face Hub (current)
- derivation = "fine_tune" — Hugging Face Hub (current)
- license = null — Hugging Face Hub (current)
- name = "Hermes" — Hugging Face Hub (current)
- base = "meta-llama/Llama-3.1-8B" — Hugging Face (fixture) (current)
- derivation = "fine_tune" — Hugging Face (fixture) (current)
- license = "llama3" — Hugging Face (fixture) (current)
- name = "Hermes 3 Llama 3.1 8B" — Hugging Face (fixture) (current)
- base = "NousResearch/Hermes-3-Llama-3.1-8B" — Hugging Face Hub (current)
- derivation = "fine_tune" — Hugging Face Hub (current)
Timeline
- Hermes 3 fine-tunes of Llama 3.1 released
- Meta releases Llama 3.1 (8B, 70B, 405B)