Skip to content
VECTOR WIREAI INTELLIGENCE
UTC

DocVQA 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 DATA26 MODELS ON THIS BOARDLEADER QWEN2 VL 72B INSTRUCT AT 96.5

The board — DocVQATOP 26 OF 26 MEASURED · BEST SCORE PER MODEL, DEFAULT VARIANT

#ModelVendorScoreVLast seen
01Qwen2 VL 72B InstructAlibaba96.52026-08-25
02Qwen2.5 VL 72BAlibaba96.42026-08-23
03MiniMax VL 01MiniMax96.42026-06-05
04InternVL2.5-78BOpenGVLab96.12026-06-05
05Qwen2.5 Omni 7BAlibaba95.22026-08-23
06Claude 3.5 SonnetAnthropic95.22026-08-23
07Qwen3 VL 4B (Reasoning)Alibaba94.92026-09-02
08Qwen2.5 VL 32B InstructAlibaba94.82026-08-26
09Llama 4 ScoutMeta94.42026-08-23
10Llama 4 MaverickMeta94.42026-09-11
11Grok 2SpaceXAI93.62026-08-23
12Nova ProAmazon93.52026-09-01
13Pixtral LargeMistral93.32026-08-23
14Grok 2 MiniSpaceXAI93.22026-08-23
15Phi 4 Multimodal InstructMicrosoft93.22026-08-23
16Gemini 2.0 Flash ExpGoogle92.92026-06-05
17GPT-4oOpenAI92.82026-08-23
18Qwen3 VL 2BAlibaba92.72026-09-02
19Nova LiteAmazon92.42026-09-01
20North-Micro-Vision-InstructCohere92.12026-08-23
21Gemini 1.5 ProGoogle91.52026-06-12
22LFM2.5-VL-3BLiquid AI91.12026-08-23
24Gemma 3 12BGoogle87.12026-08-23
25Gemma 3 27BGoogle86.62026-08-23
26Gemma 3 4BGoogle75.82026-08-23
◆ IV — INDEPENDENTLY VERIFIED · ◆ AA — AGGREGATOR · ◆ VA — VENDOR · ◇ SA — SOURCE-ATTRIBUTED

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

DocVQA-val

VARIANT19 models · best score per model
#ModelVendorScoreVLast seen
01Qwen2.5 VL 72BAlibaba95.82026-09-02
02Mage-VL-4BMicrosoft95.12026-07-26
03Qwen3.5 4BAlibaba94.82026-09-17
04Qwen3 VL 4B (Reasoning)Alibaba94.72026-07-26
05Phi 4 Multimodal InstructMicrosoft92.82026-07-26
06Qwen2.5 VL 3BAlibaba92.72026-09-02
07Qwen3.5 2BAlibaba92.62026-09-10
08North-Micro-Vision-InstructCohere92.12026-08-12
09InternVL3_5-4BOpenGVLab91.82026-09-10
10LFM2.5-VL-3BLiquid AI91.12026-08-12
11LFM2-VL-3BLiquid AI89.82026-08-22
12Ministral 3 3BMistral89.62026-08-12
13InternVL3_5-2BOpenGVLab88.42026-09-10
14LFM2.5-VL-1.6BLiquid AI87.72026-08-12
15Gemma 4 E4BGoogle87.42026-09-10

DocVQA_test

VARIANT13 models · best score per model
#ModelVendorScoreVLast seen
01Qwen3 VL 235B A22B InstructAlibaba97.12026-08-25
02Qwen3 VL 32B InstructAlibaba96.92026-08-25
03Qwen2 VL 72B InstructAlibaba96.52026-08-25
04Qwen3 VL 235B A22B ReasoningAlibaba96.52026-08-25
05Qwen3 VL 8B InstructAlibaba96.12026-08-25
06Qwen3 VL 32B ReasoningAlibaba96.12026-08-25
07Qwen3 VL Thinking (8B)Alibaba95.32026-08-25
08Qwen3 VL 4B InstructAlibaba95.32026-08-25
09Claude 3.5 SonnetAnthropic95.22026-08-24
10Qwen3 VL 30B A3B InstructAlibaba952026-08-25
11Qwen3 VL 30B A3B ReasoningAlibaba952026-08-25
12Qwen3 VL 4B (Reasoning)Alibaba94.22026-08-25
13GPT-4oOpenAI92.82026-08-24
THE FAMILY BOARD · EVERY PRINT LABELED · EVERY SCORE SOURCE-ATTRIBUTED ON ITS OWN LIVE BOARDAll leaderboardsThe frontier boardModels by vendor