Benchmark Almanac

InternVL3.5-241B-A28B

InternVL3.5-241B-A28B ranks first on 6 of the 57 benchmarks it has scores on.

Shanghai AI Lab Released 25 Aug 2025 Family InternVL3.5 The lab's model card
73
Benchmarks with a score
6
Ranked first, out of five or more models
12
In the top three
Shanghai AI Lab Aug 2025

InternVL3.5-241B-A28B

Open weights Flagship

No Epoch Capabilities Index score yet
  • ERQA 46.8% #25 /26
  • HallusionBench 57.3% #21 /23
  • BLINK 61.4% #19 /28

Where it ranks

Its best score on each benchmark, and where that ranks among every model with a score there. We rank each model once, at its best setting.

Knowledge and language

BenchmarkScoreRankSource
HellaSwag 97% #1 of 47 Model card
WinoGrande 91.7% #1 of 42 Model card
Natural Questions 50.4% #2 of 20 Model card
POPE 90.7% #1 of 7 Model card
MMLU 89.1% #20 of 127 Model card
C3 97.8% #1 of 6 Model card
CMMLU 90.2% #4 of 17 Model card
BIG-Bench Hard 86.5% #17 of 71 Model card
RACE 94.2% #1 of 4 Model card
C-Eval 90.9% #14 of 38 Model card
MMLU-Pro 81.3% #63 of 151 Model card
GAOKAO-Bench 94.5% #1 of 2 Model card
IFEval 83.7% #84 of 126 Model card
TriviaQA 74.8% #26 of 39 Model card
HallusionBench 57.3% #21 of 23 Model card

Math and science

BenchmarkScoreRankSource
MATH 94.5% #4 of 93 Model card
MATH-500 98% #9 of 67 Model card
AIME 2024 84.7% #32 of 87 Model card
GSM8K 91.6% #33 of 87 Model card
GPQA Diamond 73.2% #150 of 284 Model card
OlympiadBench 68.7% #2 of 3 Model card
AIME 2025 75.6% #101 of 143 Model card

Coding

BenchmarkScoreRankSource
HumanEval 98.2% #2 of 95 Model card

Agents and tools

BenchmarkScoreRankSource
ScreenSpot 89.8% #3 of 11 Model card
ScreenSpot-v2 92.9% #6 of 14 Model card
OSWorld-G 53.2% #4 of 5 Model card
Windows Agent Arena 18% #5 of 5 Model card

Vision, audio and robotics

BenchmarkScoreRankSource
TextVQA 84.5% #2 of 17 Model card
MM-Vet 81.2% #1 of 8 Model card
MME 2525.9 #1 of 6 Model card
RefCOCO / RefCOCO+ / RefCOCOg 92.4% #4 of 19 Model card
OCRBench 90.7% #6 of 28 Model card
MME-RealWorld 65.1% #1 of 4 Model card
VSI-Bench 69.5% #1 of 4 Model card
MMMU 77.7% #28 of 97 Model card
ChartQA 88% #12 of 37 Model card
MathVista 82.7% #21 of 64 Model card
DocVQA 94.9% #11 of 33 Model card
MMStar 77.9% #10 of 30 Model card
MVBench 76.5% #5 of 15 Model card
VCR (Visual Caption Restoration) 97% #1 of 3 Model card
OmniSpatial 51.9% #2 of 5 Model card
InfographicVQA 82% #10 of 22 Model card
VideoMME v2 (with subtitles) 23.1% #15 of 31 Official leaderboard
MTVQA-Test 39.3% #3 of 6 Model card
CRPE 78% #2 of 4 Model card
MMT-Bench 72.7% #1 of 2 Model card
R-Bench (real-world robustness) 75.4% #1 of 2 Model card
LogicVista 66.7% #7 of 12 Model card
RealWorldQA 75.2% #18 of 30 Model card
MMBench-EN v1.1 87.4% #11 of 17 Model card
AI2D 87.3% #29 of 44 Model card
MathVerse 68.5% #8 of 12 Model card
SEED-Bench-2-Plus 71.9% #2 of 3 Model card
BLINK 61.4% #19 of 28 Model card
MATH-Vision (MathVision) 63.9% #29 of 42 Model card
BabyVision 19.2% #20 of 27 Official leaderboard
MLVU 78.2% #12 of 16 Model card
Video-MME 72.9% #32 of 42 Model card
LongVideoBench 67.1% #9 of 11 Model card
DynaMath 46.5% #15 of 17 Model card
OmniDocBench v1.5 82.7 #9 of 10 Official leaderboard
We-Math 62.3% #10 of 11 Model card
MuirBench 47.5% #11 of 12 Model card
ERQA 46.8% #25 of 26 Model card
CharXiv 48.3% #65 of 67 Model card
OmniDocBench v1.6 83.8 #7 of 7 Official leaderboard
Mantis-Eval 77% #2 of 2 Model card
MMBench-Video 1.74 #2 of 2 Model card
MIRB 73% #1 of 1 Model card
MMIU 61.3% #1 of 1 Model card
SGP-Bench 70.7% #1 of 1 Model card
SpaCE-10 55% #1 of 1 Model card

Its launch post

The benchmarks the lab chose to headline, with the scores it gave.

InternVL3.5-241B-A28B model card (Hugging Face) 25 Aug 2025

Its system card

Every result in the longer documents the lab published with it. 35 of them aren't in the launch post.

InternVL3.5: Advancing Open-Source Multimodal Models in Versatility, Reasoning, and Efficiency 25 Aug 2025

Capabilities

Hallucination

Robustness