The LLM benchmark aggregator.

Llama 3.1 405B Instruct (BF16)

Meta · 405B

Strongest in Safe-prompt compliance (#41 of 82), weakest in Secure code (#238 of 273). Above par in 8 of 14 scopes. Among the models it meets almost everywhere, it finishes behind Claude Sonnet 4.5 (High, 32K, 20250929) and GPT-5 and ahead of NVIDIA-Nemotron-3-Ultra-550B-A55B (NVFP4) and GPT OSS 20B.

§ 1 · Profile

What it is good at

Bars run from 50 — the average of the models each source lists — so right of the line is above par. The middle column is the gap to whoever leads that scope.

Professional48.5−14.6#112/1881/2

Medical52.5−14#68/1491/1

Finance43.6−18.9#129/1611/1

Reasoning42.3−22.6#156/1901/5

Reasoning37.7−29#129/1441/3

Safety50.1−11.3#177/3312/3

Safe-prompt compliance53.9−6.5#41/821/1

Jailbreak resistance58.9−6.1#60/2711/1

Fairness49.4−20.6#120/2962/2

Harm refusal50.6−10.9#144/2962/2

Toxicity avoidance53.1−5.9#148/2711/1

Secure code33.1−33.2#238/2731/1

Human preference50.6−18#190/3501/1

§ 2 · Head to head

What it beats, and what beats it

The same models turn up scope after scope. Counted once: where both were placed, who finished higher. Bars run right for Llama 3.1 405B Instruct (BF16), left for the other.

Command R+ (08-2024)won 13 · lost 0

Llama 3.1 70B Instructwon 11 · lost 1

Mixtral 8x7B Instructwon 10 · lost 1

GPT-3.5 Turbowon 10 · lost 1

Jamba 1.5 Miniwon 10 · lost 1

Jamba 1.5 Largewon 10 · lost 1

Jamba 1.6 Miniwon 9 · lost 1

Jamba 1.6 Largewon 9 · lost 1

Mistral Smallwon 9 · lost 1

DeepSeek LLM 67B Chatwon 8 · lost 1

DBRX Instructwon 8 · lost 1

Llama 4 Scout Instructwon 12 · lost 2

Llama 3.1 8B Instructwon 10 · lost 2

Llama 3.3 70B Instructwon 10 · lost 2

Mixtral 8x22B Instructwon 9 · lost 2

Gemma 2 9B ITwon 7 · lost 2

Gemini 1.5 Flashwon 9 · lost 3

Command A (03-2025)won 9 · lost 3

Magistral Mediumwon 7 · lost 3

Llama 3 8B Instructwon 6 · lost 3

Mistral Largewon 8 · lost 4

Gemma 4 31Bwon 8 · lost 4

GPT-4o miniwon 9 · lost 5

Gemini 1.5 Prowon 9 · lost 5

Ling 3.0 Flashwon 6 · lost 4

GPT-4.1 nanowon 8 · lost 6

Gemini 2.0 Flashwon 8 · lost 6

Gemini 2.0 Flash-Litewon 5 · lost 4

GPT OSS 20Bwon 7 · lost 6

NVIDIA-Nemotron-3-Ultra-550B-A55B (NVFP4)won 7 · lost 6

DeepSeek V3.1won 6 · lost 6

Grok 4.3won 6 · lost 6

Grok 4.1 Fastwon 6 · lost 7

DeepSeek V4 Flashwon 6 · lost 7

Claude 3.5 Haikuwon 5 · lost 6

Claude 3 Haikuwon 4 · lost 5

Gemma 2 27B ITwon 4 · lost 5

NVIDIA Nemotron 3 Super 120B A12Bwon 4 · lost 5

Llama 4 Maverick Instructwon 6 · lost 8

DeepSeek V3won 6 · lost 8

Grok 4won 6 · lost 8

GLM-4.6won 5 · lost 7

MiMo V2.6 Flashwon 5 · lost 7

Kimi K2.6won 5 · lost 7

MiMo V2.5 Prowon 5 · lost 7

DeepSeek V3.2won 5 · lost 8

MiniMax M3won 5 · lost 8

Kimi K2.5won 5 · lost 8

Qwen3.7 Maxwon 5 · lost 8

GLM-5.2won 5 · lost 8

Gemini 3.6 Flashwon 5 · lost 8

Qwen2.5 72B Instructwon 4 · lost 7

GPT OSS 120Bwon 5 · lost 9

O3 Miniwon 5 · lost 9

Grok 3won 5 · lost 9

DeepSeek R1won 5 · lost 9

Claude 3 Sonnetwon 3 · lost 6

GLM-4.5won 4 · lost 8

Mistral Large 4won 4 · lost 9

DeepSeek-V4-Pro (High, 20260813)won 4 · lost 9

Gemini 3.5 Flashwon 4 · lost 9

Qwen3.8 Maxwon 4 · lost 9

Gemini 2.5 Flash-Litewon 3 · lost 7

GPT-4.1 miniwon 4 · lost 10

Gemini 2.5 Flashwon 4 · lost 10

Claude Haiku 4.5won 4 · lost 10

Claude 3 Opuswon 3 · lost 8

GPT-4 Turbowon 3 · lost 10

GPT-6 Lunawon 3 · lost 10

GPT-6 Solwon 3 · lost 10

GLM-5.1won 3 · lost 10

GPT-5.4won 3 · lost 10

Qwen2 72B Instructwon 2 · lost 7

Llama 3 70B Instructwon 2 · lost 7

GPT-4owon 3 · lost 11

Claude 3.5 Sonnetwon 3 · lost 11

O4 Miniwon 3 · lost 11

GPT-4.1won 3 · lost 11

Kimi K2won 3 · lost 11

Kimi K2 Thinkingwon 2 · lost 9

MiniMax M2.7won 2 · lost 10

O1 Miniwon 2 · lost 11

GPT-5.6 Lunawon 2 · lost 11

Grok 4.5won 2 · lost 11

GPT-5.6 Terrawon 2 · lost 11

GLM-5.3 Flashwon 2 · lost 11

MiMo V2.6 Prowon 2 · lost 11

GPT-5.5won 2 · lost 11

GPT-5.6 Solwon 2 · lost 11

Gemini 3.8 Flashwon 2 · lost 11

GPT-5 nanowon 2 · lost 12

Qwen3 235B A22B Instructwon 2 · lost 12

Gemini 2.5 Prowon 2 · lost 12

GPT-5.1won 2 · lost 12

MiniMax M2won 1 · lost 9

Grok 3 Miniwon 1 · lost 10

Claude Sonnet 4.6won 1 · lost 10

Gemini 3 Prowon 1 · lost 10

Muse Spark 1.1won 1 · lost 11

Inklingwon 1 · lost 12

Grok 4.6won 1 · lost 12

Claude Sonnet 5won 1 · lost 12

Claude Opus 5won 1 · lost 12

Claude 3.7 Sonnetwon 1 · lost 13

GPT-5 miniwon 1 · lost 13

Claude Sonnet 4won 1 · lost 13

O3won 1 · lost 13

Claude Opus 4.6won 0 · lost 11

GPT-5.2won 0 · lost 13

Grok 4.7won 0 · lost 13

Claude Opus 4.8won 0 · lost 13

Claude Opus 4.7won 0 · lost 13

Claude Opus 5.5won 0 · lost 13

O1won 0 · lost 14

Claude Opus 4won 0 · lost 14

GPT-5won 0 · lost 14

Claude Sonnet 4.5 (High, 32K, 20250929)won 0 · lost 14

§ 3 · Sources

Where the numbers come from

5 publications, 15 figures. Every one links to the page it was read from.

LMArena Text ↗1 measureread 2026-10-21

LMArena Text 1336

SimpleBench ↗1 measureread 2026-10-21

SimpleBench 23%

Vals.ai ↗2 measuresread 2026-10-21

Vals · TaxEval 60.88%Vals · MedQA 88.24%

HELM Safety ↗5 measuresread 2026-10-21

HELM Safety · HarmBench 62.7%HELM Safety · SimpleSafetyTests 98.8%HELM Safety · Anthropic Red Team 96.5%HELM Safety · BBQ 94.5%HELM Safety · XSTest 95.9%

Enkrypt AI Safety Leaderboard ↗6 measuresread 2026-10-21

Enkrypt · Jailbreak risk 5.3%Enkrypt · Harmful content risk 26.7%Enkrypt · CBRN risk 9.8%Enkrypt · Toxicity risk 4%Enkrypt · Bias risk 87.1%Enkrypt · Insecure code risk 67.6%

Badge

[![PublicAI Index](https://publicai.io/model-index/badge?model=llama-3-1-405b-instruct-bf16)](https://publicai.io/model-index/m/llama-3-1-405b-instruct-bf16)