Skip to content
VECTOR WIREAI INTELLIGENCE
UTC

HellaSwag leaderboard — the family board.

3 TRACKED VARIANTS · 1 FIRST-CLASS BOARD

One family, 3 tracked prints. Scoring configs and splits of the base board render as sections of this page, each under its own honest label.

● LIVE DATA48 MODELS ON THIS BOARDLEADER CLAUDE 3 OPUS AT 95.4

The board — HellaSwagTOP 48 OF 48 MEASURED · BEST SCORE PER MODEL, DEFAULT VARIANT

#ModelVendorScoreVLast seen
01Claude 3 OpusAnthropic95.42026-08-23
02GPT-4OpenAI95.32026-08-23
03Kimi K2 BaseMoonshot94.62026-06-13
04Gemini 1.5 ProGoogle93.32026-08-23
05Nemotron 3 Ultra 550B A55B BaseNVIDIA90.52026-06-05
06DeepSeek-V2.5DeepSeek90.32026-05-31
07Step3.5 Flash BaseStepFun90.22026-06-05
08GLM-4.5-BaseZ.ai90.22026-06-05
09MiMo V2.5 Pro BaseXiaomi89.82026-06-05
10MiMo V2.5 ProXiaomi89.82026-08-23
11DeepSeek-V3.2-Exp-BaseDeepSeek89.42026-06-13
12Llama 3.1 Instruct 405BMeta89.22026-05-01
13DeepSeek-V3.1-BaseDeepSeek89.22026-06-13
14Claude 3 SonnetAnthropic892026-08-23
15Nemotron 3 Super 120B A12BNVIDIA89.02026-08-11
16DeepSeek-V3DeepSeek88.92026-05-01
17Mistral Large 3 675B Base 2512Mistral88.92026-06-05
18Command R+ (Apr '24)Cohere88.62026-08-23
19MiMo V2.5 BaseXiaomi88.62026-06-05
20Hunyuan Large InstructTencent88.52026-05-31
21MiMo V2 Flash BaseXiaomi88.52026-06-13
22DeepSeek-V4-Pro-BaseDeepSeek882026-06-05
23Llama 3 (70B)Meta882026-09-10
24DeepSeek-V2DeepSeek87.82026-09-10
25Qwen2 Instruct (72B)Alibaba87.62026-08-23
26Qwen2 72BAlibaba87.62026-09-10
27Hunyuan LargeTencent86.82026-05-31
28Gemma 2 27B ITGoogle86.42026-08-23
29Claude 3 HaikuAnthropic85.92026-08-23
30DeepSeek-V4-Flash-BaseDeepSeek85.72026-06-05
31Qwen3.5 35B A3BAlibaba85.62026-08-11
32Llama 3.1 Nemotron Instruct 70BNVIDIA85.62026-08-23
33Nemotron 3.5 LightningNVIDIA85.52026-08-11
34NVIDIA Nemotron 3 Nano 30B A3BNVIDIA85.52026-08-11
35Gemma 4 26B A4BGoogle85.32026-08-11
36Qwen2.5 Instruct 72BAlibaba84.82026-05-01
37Phi 3.5 MoE InstructMicrosoft83.82026-08-23
38Mistral 7BMistral83.22026-08-29
39Qwen 2.5 Coder 32B InstructAlibaba832026-08-23
40Gemma (7B)Google82.22026-08-29
41Llama 3 Instruct 8BMeta82.12026-08-29
42Gemma 2 Instruct (9B)Google81.92026-08-23
43Phi 3 Mini 4K InstructMicrosoft75.32026-08-29
44Gemma 3n E2B InstructGoogle72.22026-09-10
45Llama 3.2 3B InstructMeta69.82026-08-23
46Phi 3.5 Mini InstructMicrosoft69.42026-08-23
47Phi 4 Mini InstructMicrosoft69.12026-08-23
48ERNIE 4.5Baidu332026-08-23
◆ IV — INDEPENDENTLY VERIFIED · ◆ AA — AGGREGATOR · ◆ VA — VENDOR · ◇ SA — SOURCE-ATTRIBUTED

Variants of this boardSCORING CONFIGS & SPLITS — NOT CITED STANDALONE, SO THEY LIVE HERE

Other tracked prints

THIN BOARDSunder 5 published models each — listed, never hidden
PrintLeaderScoreModels
HellaSwag (Acc.)Llama 3.1 Instruct 405B89.24
HellaSwag (EM)DeepSeek-V4-Pro-Base883
THE FAMILY BOARD · EVERY PRINT LABELED · EVERY SCORE SOURCE-ATTRIBUTED ON ITS OWN LIVE BOARDAll leaderboardsThe frontier boardModels by vendor