The LLM benchmark aggregator.

Llama 3.3 Instruct Turbo (70B)

Meta · 70B

Strongest in Factual grounding (#5 of 101), weakest in Secure code (#261 of 274). Above par in 4 of 15 scopes. Among the models it meets almost everywhere, it finishes behind Claude Opus 4.8 and Claude Sonnet 5 and ahead of Claude 3.5 Haiku and Gemini 1.5 Flash.

§ 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.

Professional47−14.6#114/1681/1

Legal49.2−17.9#83/1511/1

Finance42.4−20.5#123/1511/1

Knowledge38.2−25.3#121/1381/2

Academic knowledge35.8−28.1#106/1231/1

Reasoning37.4−29.5#163/1781/4

Science31.8−31.7#109/1221/1

Mathematics36.6−29.1#122/1401/2

Safety49.5−12#197/3372/3

Factual grounding64.8−5.8#5/1011/1

Jailbreak resistance56.5−8.4#94/2721/1

Toxicity avoidance55−4#110/2721/1

Harm refusal50.9−10.7#142/3001/2

Fairness42.9−27.1#246/3001/2

Secure code26−40.5#261/2741/1

§ 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 3.3 Instruct Turbo (70B), left for the other.

Command R Pluswon 12 · lost 1

Jamba 1.6 Miniwon 12 · lost 2

Jamba 1.6 Largewon 12 · lost 2

Command Awon 12 · lost 3

Mistral Smallwon 11 · lost 3

Jamba 1.5 Miniwon 7 · lost 2

Jamba 1.5 Largewon 7 · lost 2

Llama 4 Scout Instructwon 10 · lost 5

Mistral Largewon 9 · lost 5

Gemini 1.5 Flashwon 9 · lost 5

Claude 3.5 Haikuwon 8 · lost 6

GPT-4.1 Nanowon 7 · lost 7

GPT-4o Miniwon 6 · lost 7

Gemma 4 31Bwon 6 · lost 7

Mimo V2.6 Flashwon 5 · lost 6

GPT-6 Solwon 5 · lost 7

Kimi K2.6won 6 · lost 9

DeepSeek V3.2won 6 · lost 9

Kimi K2.5won 6 · lost 9

Ling 3.0 Flashwon 5 · lost 8

GPT-6 Lunawon 4 · lost 7

Gemini 1.5 Prowon 5 · lost 9

DeepSeek R1won 5 · lost 9

GLM-4.5won 5 · lost 9

DeepSeek V3won 5 · lost 10

Grok 3won 5 · lost 10

GLM-4.6won 5 · lost 10

Grok 4.1 Fastwon 5 · lost 10

GPT-5.4 Miniwon 3 · lost 7

Gemini 3.1 Flash Litewon 3 · lost 7

Gemini 3 Flashwon 3 · lost 7

MiniMax M3won 4 · lost 10

Grok 4.3won 4 · lost 10

Mimo V2.5 Prowon 4 · lost 10

GPT OSS 20Bwon 4 · lost 10

Qwen3.7 Maxwon 4 · lost 10

Gemini 3.6 Flashwon 4 · lost 10

DeepSeek V4 Flashwon 4 · lost 10

Grok 4won 4 · lost 10

Gemini 2.5 Flashwon 4 · lost 11

GPT-5 Nanowon 4 · lost 11

GPT OSS 120Bwon 4 · lost 11

Gemini 2.5 Prowon 4 · lost 11

Gemini 2.0 Flashwon 3 · lost 11

GLM-5.2won 3 · lost 11

Kimi K2 Thinkingwon 3 · lost 11

Gemini 3.5 Flashwon 3 · lost 11

GLM-5.1won 3 · lost 11

Gemini 3.8 Flashwon 3 · lost 11

Mistral Large 3won 2 · lost 8

GPT-5.4 Nanowon 2 · lost 8

GLM-5won 2 · lost 8

Claude Opus 4.1won 2 · lost 8

Grok 4 Fastwon 2 · lost 8

Claude Opus 4.5won 2 · lost 8

Qwen 3.5 Pluswon 2 · lost 8

GPT-4owon 3 · lost 12

Kimi K2won 3 · lost 12

Qwen3 235B A22B Instructwon 3 · lost 12

Claude Haiku 4.5won 3 · lost 12

DeepSeek V4 Prowon 3 · lost 12

GPT-5.5won 3 · lost 12

GPT-5.6 Solwon 3 · lost 12

Gemini 3 Prowon 2 · lost 10

GLM-5.3 Flashwon 2 · lost 12

GPT-4.1 Miniwon 2 · lost 12

O3 Miniwon 2 · lost 12

GPT-5.6 Lunawon 2 · lost 12

Grok 4.5won 2 · lost 12

MiniMax M2.7won 2 · lost 12

Qwen3.8 Maxwon 2 · lost 12

GPT-5.6 Terrawon 2 · lost 12

Llama 4 Maverick Instructwon 2 · lost 13

GPT-4.1won 2 · lost 13

O4 Miniwon 2 · lost 13

GPT-5 Miniwon 2 · lost 13

GPT-5won 2 · lost 13

GPT-5.1won 2 · lost 13

GPT-5.2won 2 · lost 13

Qwen3.5 Flashwon 1 · lost 9

Gemini 3.1 Prowon 1 · lost 9

Claude Opus 5.5won 1 · lost 10

Gemini 2.5 Flash Litewon 1 · lost 11

Claude 3.5 Sonnetwon 1 · lost 12

Claude Sonnet 4.6won 1 · lost 12

Claude Opus 4.6won 1 · lost 12

O3won 1 · lost 13

Muse Spark 1.1won 1 · lost 13

Grok 4.6won 1 · lost 13

Claude Opus 5won 1 · lost 13

Claude Opus 4won 1 · lost 14

Claude Sonnet 4won 1 · lost 14

Claude Sonnet 4.5won 1 · lost 14

Claude Opus 4.7won 1 · lost 14

GPT-5.4won 1 · lost 14

GPT-4 Turbowon 0 · lost 9

Grok 4.7won 0 · lost 11

Mimo V2.6 Prowon 0 · lost 11

Inklingwon 0 · lost 14

Claude 3.7 Sonnetwon 0 · lost 14

O1won 0 · lost 14

Claude Sonnet 5won 0 · lost 14

Claude Opus 4.8won 0 · lost 14

§ 3 · Sources

Where the numbers come from

3 publications, 12 figures. Every one links to the page it was read from.

Vals.ai ↗5 measuresread 2026-09-26

Vals · LegalBench 77.18%Vals · TaxEval 59.44%Vals · GPQA Diamond 50%Vals · MMLU Pro 69.86%Vals · AIME 16.04%

Enkrypt AI Safety Leaderboard ↗6 measuresread 2026-09-26

Enkrypt · Jailbreak risk 7.3%Enkrypt · Harmful content risk 33.9%Enkrypt · CBRN risk 12.2%Enkrypt · Toxicity risk 2.8%Enkrypt · Bias risk 88.9%Enkrypt · Insecure code risk 85.8%

Vectara Hallucination Leaderboard ↗1 measureread 2026-09-26

Vectara · Factual consistency 95.9%

Badge

[![PublicAI Index](https://publicai.io/model-index/badge?model=llama-3-3-instruct-turbo-70b)](https://publicai.io/model-index/m/llama-3-3-instruct-turbo-70b)