Request a Pilot
‹ PublicAI Index

The LLM benchmark aggregator.

GPT-4.5

OpenAI

Strongest in Harm refusal (#30 of 292, on 1 of its 2 boards), weakest in Reasoning (#154 of 195). Above par in 8 of 10 scopes. Among the models it meets almost everywhere, it finishes behind Claude Opus 5.5 and Claude Opus 4.8 and ahead of Gemini 3.6 Flash and Gemini 3.5 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.

Safety54.7#58/329
Harm refusal57.3#30/292
Safe-prompt compliance52.2#52/82
Fairness52.7#87/292
Human preference61#74/354
Human preference61#74/354
Coding50.2#86/179
Agentic coding50.3#84/173
Reasoning43.1#154/195
Reasoning40#122/147

§ 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 GPT-4.5, left for the other.

Grok Build 0.1won 6 · lost 0
DeepSeek LLM 67B Chatwon 6 · lost 0
DBRX Instructwon 6 · lost 0
Mixtral 8x7B Instructwon 6 · lost 0
Mixtral 8x22B Instructwon 6 · lost 0
Command Rwon 6 · lost 0
Llama 3.1 70B Instructwon 6 · lost 0
Gemini 2.0 Flash-Litewon 6 · lost 0
Gemini 1.5 Flashwon 7 · lost 0
Llama 3.3 70B Instructwon 7 · lost 0
QwQ 32Bwon 7 · lost 0
Gemma 3 27B ITwon 7 · lost 0
Mistral Large 3won 7 · lost 0
GPT-3.5 Turbowon 8 · lost 0
Command R+won 8 · lost 0
Gemini 1.5 Prowon 8 · lost 0
Command Awon 8 · lost 0
GPT-4o miniwon 9 · lost 1
Llama 4 Maverick Instructwon 9 · lost 1
Gemini 2.0 Flashwon 9 · lost 1
Llama 3.1 405B Instructwon 7 · lost 1
GLM-4.6won 7 · lost 1
Grok 4.3won 7 · lost 1
Claude 3 Haikuwon 6 · lost 1
Ling 3.0 Flashwon 5 · lost 1
Llama 3.1 8B Instructwon 5 · lost 1
Llama 3 8B Instructwon 5 · lost 1
Qwen1.5 72B Chatwon 5 · lost 1
OLMo 2 32B Instructwon 5 · lost 1
Mistral Small 3 24B Instructwon 5 · lost 1
Llama 3 70B Instructwon 5 · lost 1
Qwen3 Next 80B A3B Thinkingwon 5 · lost 1
Qwen3.5 Flashwon 5 · lost 1
Gemini 3.1 Flash-Litewon 5 · lost 1
GPT-4.1 nanowon 8 · lost 2
GPT-4owon 8 · lost 2
DeepSeek V3won 8 · lost 2
DeepSeek V3.2won 7 · lost 2
NVIDIA Nemotron 3 Ultra 550B A55Bwon 7 · lost 2
Grok 4.1 Fastwon 7 · lost 2
MiniMax M3won 7 · lost 2
Gemma 4 31Bwon 7 · lost 2
Llama 4 Scout Instructwon 6 · lost 2
Kimi K2.7 Codewon 5 · lost 2
Magistral Mediumwon 5 · lost 2
GPT OSS 20Bwon 5 · lost 2
MiniMax M2won 5 · lost 2
MiniMax M2.5won 5 · lost 2
Grok 4 Fastwon 5 · lost 2
DeepSeek V3.1won 5 · lost 2
GLM-4.7won 5 · lost 2
O1 Miniwon 7 · lost 3
GPT OSS 120Bwon 7 · lost 3
GPT-4.1 miniwon 7 · lost 3
Gemini 2.0 Prowon 4 · lost 2
Qwen2 72B Instructwon 4 · lost 2
Claude 3 Sonnetwon 4 · lost 2
GLM-4.5 Airwon 4 · lost 2
Inkling Smallwon 4 · lost 2
GLM-4.5won 4 · lost 2
GPT-4 Turbowon 6 · lost 3
Claude 3 Opuswon 5 · lost 3
Claude 3.5 Haikuwon 5 · lost 3
MiniMax M2.7won 5 · lost 3
GPT-4.1won 6 · lost 4
Kimi K2 Thinkingwon 4 · lost 3
Gemini 2.5 Flash-Litewon 4 · lost 3
GPT-5.4 nanowon 4 · lost 3
GPT-5.2won 5 · lost 4
DeepSeek V4 Flashwon 5 · lost 4
Kimi K2.5won 5 · lost 4
GLM-5.1won 5 · lost 4
Claude Sonnet 4.6won 5 · lost 4
Qwen3.7 Maxwon 5 · lost 4
Gemini 3.5 Flashwon 5 · lost 4
Gemini 3.6 Flashwon 5 · lost 4
Qwen2.5 72B Instructwon 3 · lost 3
Gemini 3.5 Flash-Litewon 3 · lost 3
Kimi K2.6won 4 · lost 4
MiMo V2.5 Prowon 4 · lost 4
O3 Miniwon 5 · lost 5
O4 Miniwon 5 · lost 5
Grok 3won 5 · lost 5
Grok 4won 5 · lost 5
Claude Haiku 4.5won 5 · lost 5
Qwen3 235B A22B Instructwon 5 · lost 5
Mistral Large 4won 4 · lost 5
GPT-6 Lunawon 4 · lost 5
Inklingwon 4 · lost 5
MiniMax M2.1won 3 · lost 4
GPT-5.4 miniwon 3 · lost 4
GLM-5won 3 · lost 4
Gemini 3 Flashwon 3 · lost 4
Claude Opus 4.5won 3 · lost 4
Gemini 3.1 Prowon 3 · lost 4
Grok 3 Miniwon 4 · lost 6
Claude 3.5 Sonnetwon 4 · lost 6
GPT-5 miniwon 4 · lost 6
Gemini 2.5 Flashwon 4 · lost 6
Kimi K2won 4 · lost 6
DeepSeek R1won 4 · lost 6
Gemini 2.5 Prowon 4 · lost 6
Gemini 3 Prowon 4 · lost 6
MiMo V2.6 Flashwon 3 · lost 5
Inkling 2 Smallwon 2 · lost 4
Claude Haiku 5.5won 2 · lost 4
Qwen3.8 27Bwon 2 · lost 4
Qwen3.6 Maxwon 2 · lost 4
Grok 4.20won 2 · lost 4
Grok 4.7won 3 · lost 6
GPT-5.6 Lunawon 3 · lost 6
DeepSeek V4 Prowon 3 · lost 6
GLM-5.2won 3 · lost 6
MiMo V2.6 Prowon 3 · lost 6
Qwen3.8 Maxwon 3 · lost 6
Claude 3.7 Sonnetwon 3 · lost 7
Claude Sonnet 4won 3 · lost 7
Claude Opus 4won 3 · lost 7
GPT-5 nanowon 2 · lost 6
GPT-6 Solwon 2 · lost 7
GLM-5.3 Flashwon 2 · lost 7
GPT-5.4won 2 · lost 7
GPT-5.6 Solwon 2 · lost 7
Gemini 3.8 Flashwon 2 · lost 7
Claude Opus 4.6won 2 · lost 7
O1won 2 · lost 8
O3won 2 · lost 8
GPT-5won 2 · lost 8
GPT-5.1won 2 · lost 8
Claude Sonnet 4.5won 2 · lost 8
GPT-6 Astrawon 1 · lost 6
DeepSeek V4.1 Flashwon 1 · lost 6
Gemini 4 Argonwon 1 · lost 6
Grok 4.6won 1 · lost 8
Grok 4.5won 1 · lost 8
GPT-5.6 Terrawon 1 · lost 8
GPT-5.5won 1 · lost 8
Claude Opus 5won 1 · lost 8
Claude Opus 4.7won 1 · lost 8
Step 5won 0 · lost 6
Claude Sonnet 5.5won 0 · lost 6
GPT-6.1 Solwon 0 · lost 6
Kimi K3.1 Prowon 0 · lost 6
GLM-5.3won 0 · lost 6
Kimi K3won 0 · lost 6
Gemini 3.7 Flashwon 0 · lost 6
Muse Spark 1.2won 0 · lost 6
Muse Spark 1.3won 0 · lost 6
Claude Fable 5.1won 0 · lost 6
Claude Fable 5won 0 · lost 6
Muse Wonderwon 0 · lost 6
Gemini 4 Carbonwon 0 · lost 6
Muse Spark 1.1won 0 · lost 8
Claude Sonnet 5won 0 · lost 9
Claude Opus 4.8won 0 · lost 9
Claude Opus 5.5won 0 · lost 9

§ 3 · Sources

Where the numbers come from

5 publications, 9 figures. Every one links to the page it was read from.

LMArena Text ↗1 measureread 2026-11-10
LMArena Text 1445
ARC-AGI-2 ↗1 measureread 2026-11-10
ARC-AGI-2 0.8%
Aider polyglot ↗1 measureread 2026-11-10
Aider polyglot 44.9%
SimpleBench ↗1 measureread 2026-11-10
SimpleBench 34.5%
HELM Safety ↗5 measuresread 2026-11-10
HELM Safety · HarmBench 95.8%HELM Safety · SimpleSafetyTests 100%HELM Safety · Anthropic Red Team 99.4%HELM Safety · BBQ 92%HELM Safety · XSTest 95.1%

Badge

PublicAI Index badge for GPT-4.5[![PublicAI Index](https://publicai.io/model-index/badge?model=gpt-4-5)](https://publicai.io/model-index/m/gpt-4-5)