Multimodal AI
Multimodal AI is moving from interpreting media to carrying out work with it. Qwen's September report describes audio and video agents and supporting tools. Available models still differ in which inputs and outputs they support. Benchmark results, native architecture and reliable real-time operation each need their own test.
Research snapshot · updated Nov 12, 2026
Core thesis
Native multimodality is the default architecture for frontier models, not a frontier in itself: the leading text-image-audio-video stacks (Gemini, GPT-5.x, Claude) ship multimodal by default, and Google's Gemini line carries a 2M-token context that ingests hours of video. The live contest is real-time omni interaction, on-device inference and agentic computer use. Enterprise value concentrates in document understanding, visual inspection and video analysis. The constraint is inference economics: multimodal still costs several times more per query than text, and the firms that drive that cost down fastest capture the margin.
State of the art (2026)
Leading models now take several kinds of input in a single model. Google's documentation for Gemini 2.5 Pro, for example, lists text, image, audio, video and PDF inputs, a limit of 1,048,576 input tokens and text output of up to 65,536 tokens. Models still differ in what they accept and what they can produce. Competition has moved on to real-time voice and video conversation, running models on devices, and agents that operate a computer.
Explore the deeper analysis in CanaryIQ
Explore what is covered below. These previews show the structure; log in to read the full analysis.
Signal stack
Evidence stacked leading → lagging
10 signals
Signal stack
Evidence stacked leading → lagging
Technology-native KPIs
Metrics that predict trajectory, tracked over time
3 tracked
Technology-native KPIs
Metrics that predict trajectory, tracked over time
Landscape map
Who builds what, and who depends on whom
139 players · 6 layers
Landscape map
Who builds what, and who depends on whom
Catalyst calendar
Dated events that will move the position
6 ahead
Catalyst calendar
Dated events that will move the position
Technology roadmap
Milestones on the path to maturity
8 milestones
Technology roadmap
Milestones on the path to maturity
Watchlists
Companies, people and papers, each with a condition for removing it
20 · 20
Watchlists
Companies, people and papers, each with a condition for removing it
Decision frameworks
What our view means for your job
Locked
Decision frameworks
What our view means for your job
Thesis changelog
When our view changed, and why
8 updates
Thesis changelog
When our view changed, and why
Change our mind
2 disconfirming conditions
You've read the verdict. The file is much deeper.
The full signal stack, technology-native KPIs tracked over time, the landscape of who depends on whom, the dated catalyst calendar, decision frameworks for every desk, live watchlists and a record of every time our view on Multimodal AI has changed. All of it is in CanaryIQ.