Skip to content
VECTOR WIREAI INTELLIGENCE
UTC

GSM8K leaderboard — the family board.

4 TRACKED VARIANTS · 1 FIRST-CLASS BOARD

One family, 4 tracked prints. Scoring configs and splits of the base board render as sections of this page, each under its own honest label.

● LIVE DATA142 MODELS ON THIS BOARDLEADER MIMO V2.5 PRO AT 99.6

The board — GSM8KTOP 100 OF 142 MEASURED · BEST SCORE PER MODEL, DEFAULT VARIANT

#ModelVendorScoreVLast seen
01MiMo V2.5 ProXiaomi99.62026-08-23
02MiMo V2.5 Pro BaseXiaomi99.62026-06-05
03Kimi K2 InstructMoonshot97.32026-08-23
04o1OpenAI97.12026-08-23
05GPT-4.5 PreviewOpenAI972026-08-23
06Claude 3.5 SonnetAnthropic96.92026-08-23
07Llama 3.1 Instruct 405BMeta96.82026-08-23
08DeepSeek-V3DeepSeek96.72026-07-02
09Gemma 3 27BGoogle95.92026-08-23
10Qwen2.5 Instruct 72BAlibaba95.82026-08-23
11GPT-4oOpenAI95.62026-07-02
12Gemini 2.0 Flash ExpGoogle95.42026-06-06
13Hy3-preview-BaseTencent95.42026-06-05
14Gemini 1.5 ProGoogle95.22026-08-23
15DeepSeek-V2.5DeepSeek95.12026-08-23
16Claude 3 OpusAnthropic952026-08-23
17Nova ProAmazon94.82026-09-01
18MiniMax Text 01MiniMax94.82026-06-06
19Nova LiteAmazon94.52026-09-01
20Gemma 3 12BGoogle94.42026-08-23
21Qwen3 235B A22BAlibaba94.42026-08-23
22Llama 3.3 70B InstructMeta94.22026-07-02
23Granite 4.1 30BIBM94.22026-06-15
24Llama 3.1 70B InstructMeta93.82026-07-02
26Kimi K2 BaseMoonshot93.52026-08-23
27GPT-4OpenAI93.22026-08-12
28Mistral Large 2Mistral932026-08-23
29Hunyuan LargeTencent92.82026-05-31
30DeepSeek-V4-Pro-BaseDeepSeek92.62026-06-05
31Granite 4.1 8BIBM92.52026-06-15
32Claude 3 SonnetAnthropic92.32026-08-23
33Nova MicroAmazon92.32026-09-01
34MiMo V2 Flash BaseXiaomi92.32026-06-13
35Qwen2 Instruct (72B)Alibaba922026-08-23
36NVIDIA Nemotron 3 Nano 30B A3BNVIDIA91.42026-08-11
37Llama 3.1 Nemotron Instruct 70BNVIDIA91.42026-08-23
38DeepSeek-V3.1-BaseDeepSeek91.42026-06-13
39Nemotron 3.5 LightningNVIDIA91.32026-08-11
40Mistral Large 3 675B Base 2512Mistral91.22026-06-05
41DeepSeek-V3.2-Exp-BaseDeepSeek91.12026-06-13
42Qwen 2.5 Coder 32B InstructAlibaba91.12026-08-23
43DeepSeek-V4-Flash-BaseDeepSeek90.82026-06-05
44Nemotron 3 Super 120B A12BNVIDIA90.72026-08-11
45Palmyra-X-004Writer90.52026-07-02
46Qwen3.5 35B A3BAlibaba90.12026-08-11
47GLM-4.5-BaseZ.ai90.12026-06-05
48Gemma 3 4BGoogle89.92026-10-07
49Qwen2 72BAlibaba89.52026-09-10
50Claude 3 HaikuAnthropic88.92026-08-23
51Qwen2.5 Omni 7BAlibaba88.72026-08-23
52Phi 3.5 MoE InstructMicrosoft88.72026-08-23
53Phi 4 Mini InstructMicrosoft88.62026-08-23
54Step3.5 Flash BaseStepFun88.22026-06-05
55DeepSeek-V3-BaseDeepSeek88.22026-06-05
56Nemotron 3 Ultra 550B A55B BaseNVIDIA88.12026-06-05
57Claude 2Anthropic882026-08-12
58Phi 3 Medium 4K InstructMicrosoft87.82026-07-03
59Solar ProUpstage87.12026-07-02
60Jamba 1.5 LargeAI21872026-08-23
61Granite 4.1 3BIBM86.92026-06-15
62Phi 3.5 Mini InstructMicrosoft86.22026-08-23
63LFM2.5-1.2B-ThinkingLiquid AI85.62026-08-09
64Qwen3 1.7BAlibaba85.62026-10-07
65Claude 1.3Anthropic85.22026-07-17
66GPT-4o miniOpenAI84.32026-07-02
67MAI-Thinking-1Microsoft842026-08-14
68Granite 4.1 30B BaseIBM83.82026-06-15
69MiMo V2.5 BaseXiaomi83.32026-06-05
70Llama 3.1 8B InstructMeta83.22026-07-02
71Palmyra X V3 (72B)Writer83.12026-07-02
72Llama 3 (70B)Meta832026-09-10
73Qwen2.5 Instruct Turbo (7B)Alibaba832026-07-02
74Granite 4.0 H TinyIBM82.62026-10-07
75LFM2-2.6BLiquid AI82.42026-10-07
76GPT-4 TurboOpenAI82.42026-07-02
77Qwen2 7B InstructAlibaba82.32026-08-23
79Granite 3.2 8B InstructIBM81.72026-06-11
80Claude 3.5 HaikuAnthropic81.52026-07-02
81Gemma 2 27B ITGoogle81.22026-08-23
82Granite 3.3 8B InstructIBM80.92026-08-23
83DeepSeek-LLM-67B-Chat (V1)DeepSeek79.52026-07-02
84DeepSeek-V2DeepSeek79.22026-09-10
85Granite 3.1 8B InstructIBM79.22026-06-11
86Qwen2.5 3BAlibaba79.12026-06-05
87Gemini 1.5 Flash (001)Google78.52026-07-02
88DeepSeek-R1-Distill-Qwen-7BDeepSeek78.52026-06-11
89Mistral NemoMistral78.22026-07-02
90Llama 3.2 3B InstructMeta77.72026-10-07
91MoonlightMoonshot77.42026-06-12
92Gemma 4 26B A4BGoogle77.02026-08-11
93Arctic InstructSnowflake76.82026-07-02
94Gemma 2 Instruct (9B)Google76.22026-08-23
95Jamba 1.5 MiniAI2175.82026-08-23
96Command R+ (Apr '24)Cohere73.82026-08-23
97Granite 4.1 8B BaseIBM73.52026-06-15
98Palmyra X V2 (33B)Writer73.52026-07-02
99Granite 4.0 1BIBM73.42026-08-09
100Mistral Small (2402)Mistral73.42026-07-02
◆ IV — INDEPENDENTLY VERIFIED · ◆ AA — AGGREGATOR · ◆ VA — VENDOR · ◇ SA — SOURCE-ATTRIBUTED

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

GSM8K (EM)

VARIANT11 models · best score per model
#ModelVendorScoreVLast seen
01DeepSeek-V4-Pro-BaseDeepSeek92.62026-09-10
02DeepSeek-V3.2-BaseDeepSeek91.12026-06-27
03DeepSeek-V4-Flash-BaseDeepSeek90.82026-09-10
04DeepSeek-V3DeepSeek89.32026-05-01
05Qwen2.5 Instruct 72BAlibaba88.32026-05-01
06Llama 3.1 Instruct 405BMeta83.52026-05-01
07DeepSeek-V2DeepSeek81.62026-05-01
08Qwen2.5 7B BaseAlibaba36.42026-05-03
09Qwen2.5 3B BaseAlibaba29.22026-05-03
10Qwen3 4B Base DenseAlibaba28.32026-05-03
11Gemma 3 1B Base DenseGoogle9.72026-05-03

Other tracked prints

THIN BOARDSunder 5 published models each — listed, never hidden
PrintLeaderScoreModels

Also tracked, no published surface rows yet: GSM8k* (speech understanding)

THE FAMILY BOARD · EVERY PRINT LABELED · EVERY SCORE SOURCE-ATTRIBUTED ON ITS OWN LIVE BOARDAll leaderboardsThe frontier boardModels by vendor