Llama-3.1-405B-Instruct
Meta Llama- Arena score
- 1284
- Among open models
- #92of 148
- Among all on Arena
- #270
- Votes
- 41,375
Measured on Llama 3.1 405B Instruct (BF16), the served version, not a local build. LMArena, CC BY 4.0, published Nov 11, 2026.
Score by category
CategoryScoreOpen rankArena rankVotes
- English1312#89#25923.3K
- Legal & government1303#86#2513.4K
- Multi-turn1297#82#2526.8K
- Science1295#89#2637.2K
- Hard prompts, English1291#91#2666.8K
- Coding1291#91#2686.2K
- Software & IT1290#95#27410.3K
- Medicine & health1286#89#2581.9K
- Math1281#86#2548.5K
- French1279#73#230364
- Applied math1276#88#2647.2K
- Writing & literature1274#86#26516K
- Hard prompts1269#95#27410.7K
- Long prompts1266#94#2805.5K
- Russian1265#86#2667.7K
- Creative writing1262#86#2638.9K
- Spanish1260#79#241348
- Instructions1258#90#26916.2K
- German1252#75#235894
- Media & sports1250#87#2659.5K
- Non-English1248#96#28327.2K
- Business & finance1247#97#2796.7K
- Expert1243#93#2673.1K
- Chinese1242#93#2793.1K
- Excluding ties1213#92#26927.1K
- Japanese1208#60#211846
- Korean1184#72#234924
| Model | Params | Score |
|---|---|---|
| 8.0B | — | |
| 8.0B | 1187 | |
| 70.6B | 1261 | |
| 406B | 1284 |
About as strong, smaller
Compare
granite-4.2-3bIBM Granite12923.7B
gemma-3-4b-itGoogle12914.3B

gemma-3n-E4B-itGoogle13067.8B

Molmo2-8BAi212878.7B
Discussion
- Show HN: Llama 3.1 70B on a single RTX 3090 via NVMe-to-GPU bypassing the CPUxaskasdf · Feb 21, 2026395 points
- Comparativa LLM Local: Llama 3.1 vs Gemma 4 vs Qwen 3.8 ¿Cuál es el MEJOR?Ricardo Bertran · Sep 7, 2026205 likes
- Llama 3.1 8B ⚠️ ¿Sigue siendo VALIOSO o ya es OBSOLETO?Nichonauta · Sep 2, 202682 likes
- 🔥 Samsung научила память считать: Llama 3.1 ускорилась в 3 раза без новой GPU #ai #shortsAI Практика Pro · Aug 26, 202638 likes
- Build Your Own Local AI Chatbot on Mac | Open WebUI + Ollama + Llama 3.1ProgramGuru · Sep 17, 202622 likes
Built on it
Ways to run it
- GGUF · llama.cpp, Ollama, LM StudioMaziyarPanahi/Meta-Llama-3.1-405B-Instruct-GGUF4 builds
- MLX · Apple silicon: mlx-lm, LM Studiomlx-community/Meta-Llama-3.1-405B-Instruct-8bit3 builds
- AWQ · vLLM, SGLangkishizaki-sci/Llama-3.1-405B-Instruct-AWQ-4bit-JP-EN1 build
- GPTQ · vLLM, SGLangxmadai/Llama-3.1-405B-Instruct-xMADai-INT41 build
Most downloaded
- Meta-Llama-3.1-405B-Instruct-GGUFMaziyarPanahi · Quantization36K downloads
- Llama-3.1-405B-Instruct-NVFP4nvidia · Fine-tune3.4K downloads
- Meta-Llama-3.1-405B-Instruct-FP8-dynamicRedHatAI · Quantization2.9K downloads
- Llama-3.1-405B-Instruct-FP8nvidia · Fine-tune2.5K downloads
- Meta-Llama-3.1-405B-Instruct-FP8RedHatAI · Quantization980 downloads
BenchClub runs
Not runnable yet
Runs open once a Llama-3.1-405B-Instruct build is registered.
More from Meta Llama
See more
Llama-3.1-8B-Instruct4.2M8.0B

Llama-3.2-1B-Instruct1.3M1.2B

Llama-3.2-3B-Instruct831.4K3.2B

Llama-3.2-1B644.1K1.2B
Other >300B models
See more
DeepSeek-V3.2DeepSeek9.6M685B
GLM-5.3-FlashZ.ai6.4M321B
InklingThinking Machines1.5M952B
NVIDIA-Nemotron-3-Ultra-550B-A55B-BF16NVIDIA1M561B