VECTOR WIREAI INTELLIGENCE
⌘K
NVDA$1,847+3.2%MSFT$512+1.1%GOOGL$199-0.4%META$728+2.7%AMD$184-1.2%TSM$212+0.6%PLTR$98+4.1%AI IDX4,821+1.9%
PKT
SEEDRefresh Models —Deals —Regulatory —Sources —

harmbench

73 models tracked

HarmBench (safety/harmfulness refusal benchmark) — Standard published safety benchmark via HELM.

#ModelVendorBest scoreRunsLast seen
1GPT OSS 120BOpenAI10012026-05-10
2Llama 3.1 70B InstructMeta10012026-07-05
3Qwen3 8BAlibaba10012026-07-05
4o3OpenAI98.412026-05-10
5Claude 3.5 SonnetAnthropic98.112026-05-10
6Claude Sonnet 4Anthropic9832026-06-21
7Claude 3 OpusAnthropic97.412026-05-10
8Kimi K2 InstructMoonshot97.412026-05-10
9o4-miniOpenAI9712026-05-10
10o1OpenAI96.322026-06-18
11GPT-4.5 PreviewOpenAI95.812026-05-10
12Claude 3 SonnetAnthropic95.812026-05-10
13o3-miniOpenAI95.212026-05-10
14Claude 4 OpusAnthropic91.722026-06-18
15GPT-4.1OpenAI91.712026-05-10
16Claude 3 HaikuAnthropic91.312026-05-10
17GPT-4 TurboOpenAI89.812026-05-10
18OpenAI o1-miniOpenAI88.512026-05-10
19Shieldstral-1.0-3BMistral8712026-08-04
20GPT-4.1 nanoOpenAI86.822026-06-19
21GPT-4.1 miniOpenAI85.622026-06-19
22Nemotron 3.5 Content Safety 4BNVIDIA85.312026-08-04
23GPT-4o miniOpenAI84.912026-05-10
24Claude 3.7 SonnetAnthropic84.322026-06-21
25OLMo 2 32B Instruct March 2025AllenAI84.112026-05-10
26Palmyra FinWriter84.112026-05-10
27GPT-4oOpenAI82.912026-05-10
28Granite 3.3 8B InstructIBM81.212026-05-10
29Gemini 1.5 Flash (001)Google8012026-05-10
30Gemini 1.5 ProGoogle79.912026-05-10
31Qwen2 Instruct (72B)Alibaba76.812026-05-10
32Palmyra X5Writer73.812026-05-10
33Palmyra-X-004Writer73.812026-05-10
34Qwen2.5 Instruct 72BAlibaba72.812026-05-10
35Llama 3 Instruct 8BMeta72.722026-06-19
36Gemini 2.0 Flash (Non-Reasoning)Google72.212026-05-10
37Gemini 2.0 Flash Lite (02-05 preview)Google72.212026-06-24
38Gemini 3 Flash PreviewGoogle7012026-07-05
39GPT-5OpenAI7012026-07-05
40Qwen2.5 Instruct Turbo (7B)Alibaba67.712026-05-10
41Gemini 2.5 Flash LiteGoogle6712026-05-10
42Gemini 2.0 Flash (Reasoning)Google66.212026-06-24
43Llama 4 Maverick 17B 128E Instruct FP8Meta66.112026-05-10
44Gemini 2.5 ProGoogle65.412026-05-10
45Gemini 2.0 Pro (02-05 preview)Google65.312026-05-10
46DeepSeek-ChatDeepSeek64.912026-05-10
47Llama 3 Instruct 70BMeta6422026-06-19
48OLMoE 1B-7B Instruct January 2025AllenAI62.912026-05-10
49Llama 3.1 Instruct 405BMeta62.712026-05-10
50Gemini 2.5 FlashGoogle62.612026-05-10
51Llama 3.1 Instruct Turbo (8B)Meta61.612026-05-10
52Llama 4 Scout (17Bx16E) InstructMeta6012026-06-26
53Llama 4 ScoutMeta6012026-05-10
54OLMo 2 7B Instruct November 2024AllenAI57.812026-05-10
55Grok 3 Mini BetaSpaceXAI57.212026-05-10
56Qwen3 235B A22B FP8 ThroughputAlibaba56.912026-05-10
57GLM-4.5-AirZ.ai56.112026-05-10
58Palmyra MedWriter55.612026-05-10
59DeepSeek-R1DeepSeek54.622026-05-10
60OLMo 2 13B Instruct November 2024AllenAI51.912026-05-10
61Command RCohere50.412026-05-10
62DeepSeek-V3DeepSeek49.712026-05-10
63Command R+ (Apr '24)Cohere48.512026-05-10
64Llama 3.1 Instruct Turbo (70B)Meta46.912026-05-10
65Grok 3 BetaSpaceXAI45.312026-05-10
66Mistral Small 3 (Non-Reasoning)Mistral45.222026-06-19
67Grok 4SpaceXAI39.712026-05-10
68gpt-oss-20bOpenAI39.512026-07-05
69Mistral Instruct V0.3 (7B)Mistral34.412026-05-10
70DBRX InstructDatabricks27.112026-05-10
71Claude Sonnet 4.6Anthropic24.212026-07-05
72Mistral V0.1 (7B)Mistral23.312026-05-10
73Claude Opus 4.7Anthropic17.512026-07-05

Best tracked score per model (default configuration; source-attributed and verification-tiered). Open a model for its full benchmark surface, provenance and pricing.