Skip to content

VLM Intelligence Index

Overshoot VLM Intelligence Index

Composite vision score across nine evals, run on live video. Higher is better.

Overshoot VLM Intelligence Index. 13 bars. Highest: Claude Opus 4.6 at 86. Toggle the data table for exact values.02040608010086AN85OA83AN83GO81AL77AL76OA75GO74AN74HC73GO69GO64OA

Reasoning models are indicated by a lightbulb icon

#Details
1View Claude Opus 4.6Claude Opus 4.6maxAnthropic
86
Closed400k
2View GPT-5.4GPT-5.4xhighOpenAI
85
Closed400k
3View Claude Sonnet 4.6Claude Sonnet 4.6xhighAnthropic
83
Closed300k
4View Gemini 3.1 ProGemini 3.1 ProxhighGoogle
83
Closed2M
5View Qwen3.6 35B A3BQwen3.6 35B A3BxhighAlibaba (Qwen)
81
Open35B256k
6View Qwen3.6 27BQwen3.6 27BAlibaba (Qwen)
77
Open27B256k
7View GPT-5.4 miniGPT-5.4 minixhighOpenAI
76
Closed256k
8View Gemini 3 FlashGemini 3 FlashGoogle
75
Closed1M
9View Claude Haiku 4.5Claude Haiku 4.5Anthropic
74
Closed250k
10View Holo3 35B A3BHolo3 35B A3BH Company
74
Open35B128k
11View Gemma 4 31BGemma 4 31BGoogle
73
Open31B128k
12View Gemma 4 26B A4BGemma 4 26B A4BGoogle
69
Open26B128k
13View GPT-5.4 nanoGPT-5.4 nanoxhighOpenAI
64
Closed128k