C
ChaoBro

Google I/O Preview Leaks: Gemini "Omni" Multimodal Model + 3.5 Flash + New Vision Model, Triple Release Warmup

Google I/O Preview Leaks: Gemini "Omni" Multimodal Model + 3.5 Flash + New Vision Model, Triple Release Warmup

Core Judgment

With Google I/O just days away, leak information about Gemini’s product line is pouring in. The core signal: Google is no longer satisfied with being “a better chatbot” — it wants to build full-scenario AI infrastructure covering text, video, and vision.

The leaks involve three product lines, each targeting a different market position — this is not a single model upgrade, but a strategic-level product matrix restructuring.

Three Leaked Product Lines

1. Gemini “Omni” Multimodal Model

AttributeInformation
PositioningUltra-deep multimodal understanding and generation
Key CapabilityBeyond current Veo’s video generation quality, supporting bidirectional video understanding + generation
Current StatusInternal testing, UI already shows “powered by Omni” label
Release WindowDuring or shortly after Google I/O

Omni’s core value lies in unifying understanding and generation. Current AI models are typically one-way — they can understand video but not generate it, or generate but not understand. If Omni achieves bidirectional capability, it will become the first truly “omni-modal” model.

Leaked UI screenshots show a “powered by Omni” label, indicating Google plans to integrate it as an underlying engine across multiple products, rather than as a standalone chat interface.

2. Gemini 3.5 Flash

AttributeInformation
PositioningHigh-speed, low-cost everyday reasoning model
Current StatusAlready in internal testing
Expected ReleaseGoogle I/O
Competitor TargetGPT-4o mini, Claude Haiku

3.5 Flash continues Google’s “Flash” series positioning — not pursuing the strongest intelligence, but pursuing the fastest response speed and lowest cost. For enterprise users needing large-scale AI deployment, this is the most pragmatic choice.

3. “spark Robin” Vision Model

AttributeInformation
PositioningNew model focused on image/vision understanding
Current StatusLeaked stage, limited details
Potential UseGoogle Lens upgrade, Photos smart search, Android system-level vision

The “spark Robin” naming suggests it belongs to Google’s “Spark” model series (Muse Spark is the flagship of this series). If this is a standalone vision model, Google may integrate it into the Android system, achieving system-level AI vision capability.

Google I/O 2026 Potential Full Picture

ProductPositioningTarget Users
Gemini 3.5 FlashHigh-speed low-cost reasoningDevelopers, enterprise batch deployment
Gemini OmniFull-modal understanding + generationPremium users, creative industries
spark RobinVision-specific modelMobile devices, system integration
Daily BriefAI daily briefingIndividual users
Cosmo AI (Nano)On-device AI applicationMobile devices
AI AvatarsDigital humansSocial, customer service scenarios

Competitor Landscape Comparison

CompanyMultimodal StrategyCurrent Strongest
GoogleOmni unifies understanding + generationGemini 2.5 Pro
OpenAIGPT-4o multimodal + Veo videoGPT-5.5
AnthropicClaude native multimodalClaude 5 (Mythos)
AlibabaQwen-VL + Tongyi WanxiangQwen3.6-Max

Google’s Omni strategy is most similar to OpenAI’s GPT-4o — both pursuing a single model that handles all modalities. But unlike OpenAI’s “one super-large model” approach, Google has chosen a multi-model matrix strategy: Flash for volume, Omni for heavy lifting, Robin for specialization. This strategy’s advantage is flexibility and cost control; the disadvantage is higher ecosystem integration difficulty.

Action Recommendations

For Developers

  • Watch Flash 3.5 API pricing: If it continues the Flash series’ low-price strategy, it could be the best choice for batch deployment
  • Evaluate Omni’s video capabilities: If Veo-level video generation is available via API, it will dramatically lower the barrier for video content production
  • Prepare multi-model routing: Google’s multi-model matrix means you’ll need smart routing strategies to choose the right model

For Enterprises

  • Google ecosystem users get priority: Companies already using Google Workspace will experience Gemini’s deep integration first
  • Video content producers: Omni’s video generation capability may change video content production workflows
  • Mobile developers: Cosmo AI (Nano)‘s on-device capabilities are worth watching, especially for privacy-sensitive scenarios

Risk Reminder

All current information comes from leaks and is not officially confirmed. Google I/O’s actual release content may differ from leaked information. Historically, Google has also leaked multiple pieces of information before I/O that never materialized. Await official announcements.