Google releases Gemini 3.5 Flash with 4x faster output and agentic capabilities, 3.5 Pro coming June
Google released Gemini 3.5 Flash today with 4x faster output token generation than competing frontier models while surpassing Gemini 3.1 Pro on coding, agentic, and multimodal benchmarks. The company announced Gemini 3.5 Pro will launch next month and introduced Gemini Omni, a new multimodal series that outputs video.
Gemini 3.5 Flash at a glance
How Gemini 3.5 Flash ranks now
| Benchmark | Score | Rank | Source |
|---|---|---|---|
| Arena Elo | 1478 Elo | 26 of 141 | LMArena, Nov 2026 ↗ |
| GPQA | 92.2% | 25 of 164 | Artificial Analysis, Jul 2026 ↗ |
| Speed (tok/s) | 167 tok/s | 17 of 103 | Artificial Analysis, Nov 2026 ↗ |
Rank is against every model in our database with a sourced score on that benchmark, each counted once with its best score.
Google releases Gemini 3.5 Flash with 4x faster output and agentic capabilities, 3.5 Pro coming June
Google released Gemini 3.5 Flash today at I/O 2026, claiming the model delivers 4x faster output tokens per second compared to other frontier models while maintaining Flash series pricing. According to Google, the model surpasses Gemini 3.1 Pro in coding, agentic, and multimodal benchmarks.
The model is available immediately in the Gemini app, Google Search, Antigravity 2.0, and via the Gemini API. Gemini 3.5 Pro is currently in testing and will launch next month.
New Gemini Omni multimodal series
Google introduced Gemini Omni, a new model family combining reasoning with content creation. Gemini Omni Flash accepts image, audio, video, and text inputs and outputs video "grounded in real-world knowledge" that can be edited. The model is rolling out to AI Plus, Pro, and Ultra subscribers in the Gemini app, Google Flow, and YouTube Shorts.
The company also launched Google Flow and Flow Music as standalone mobile apps, with Flow available on Android in beta and Flow Music launching on iOS first.
Gemini Spark agentic assistant
Google announced Gemini Spark, described as "your personal agent" that performs tasks autonomously. The system integrates with Gmail, Docs, and other Google Workspace apps, with third-party tool support via MCP (Model Context Protocol) coming this summer.
Gemini Spark will be available next week exclusively to Google AI Ultra subscribers in the US. The service transforms Gemini "from an assistant that can answer your questions into an active partner that does real work on your behalf," according to Google.
Pricing changes for Google AI plans
Google restructured its subscription tiers. AI Ultra now starts at $100 per month (previously $250), offering 5x higher usage limits than AI Pro. A new $200 tier replaces the previous $250 plan with identical capabilities.
The company is shifting from daily prompt limits to a "compute-used" model that accounts for prompt complexity, features used, and chat length. Limits refresh every five hours until reaching a weekly cap.
Search and productivity updates
AI Mode in Google Search now runs on Gemini 3.5 Flash. New "information agents" will monitor the web 24/7 for topics users specify, available to AI Pro and Ultra subscribers this summer. Google Search will also gain the ability to build custom dashboards and trackers for ongoing tasks.
Gmail Live, a conversational email search feature, rolls out to AI Pro and Ultra subscribers in the US this summer on Android and iOS. Docs Live for conversational document creation and editing launches simultaneously for the same subscriber tiers.
What this means
Google's 4x speed claim for Gemini 3.5 Flash positions it directly against Anthropic's Claude 3.5 Sonnet and OpenAI's GPT-4o in the fast inference tier. The introduction of Gemini Omni's video output capabilities represents a significant multimodal expansion, though real-world quality benchmarks remain to be seen. The restructured pricing and compute-based limits suggest Google is attempting to balance access with infrastructure costs as model capabilities increase. The MCP integration for Gemini Spark indicates Google is adopting industry standards for agent interoperability rather than building a proprietary ecosystem.
Related Articles
RadixArk releases DaisyRose-27B, lifting Terminal-Bench 4.1 pass@1 from 5.6% to 12.0%
RadixArk released DaisyRose-27B, its first model, built on Qwen3.8-27B and trained with reinforcement learning on 1,200 curated agent tasks. According to RadixArk, it raises Terminal Bench 4.1 pass@1 from 5.6% to 12.0% and scores higher than the base model on all seven reported benchmarks. Weights ship under Apache 2.0 in BF16, FP8 and NVFP4.
Google releases Genie 4 to AI Ultra subscribers with real-time prompt editing and native audio
Google is rolling out Genie 4, which it calls its most realistic and interactive world model yet, to eligible Google AI Ultra subscribers in Project Genie. The update adds real-time native audio, mid-session text prompting, a world-building agent and a community gallery. Genie 3 also gains sessions of up to 3 minutes.
Tencent releases Hy4 flash: 220B MoE, 12B active parameters, 1M-token context
Tencent has released Hy4 flash, a multimodal Mixture-of-Experts model with 220B total parameters, 12B active per token, and a 1,048,576-token context window. Tencent calls it its most cost-effective Hy model to date and claims a per-task cost of 1/20 that of Hy4 on its internal WorkIndex benchmark.
Kuaishou launches Kling Image 4.0 with native 8K image generation and up to 10 images per prompt
Kuaishou launched Kling Image 4.0 on Tuesday night, adding native 8K image generation and up to 10 coordinated images from a single prompt. The model is available through an API for developers and enterprise customers.
Comments
Loading...