AI News Briefing

News MarkTechPost May 20

Google Introduces Gemini 3.5 Flash at I/O 2026: A Faster and Cheaper Model for AI Agents and Coding

By Michal Sutter

46 score

AI Analysis

Continuing our coverage from yesterday, Google released Gemini 3.5 Flash at I/O 2026, its first Gemini 3.5 model. It outperforms Gemini 3.1 Pro on benchmarks, is 4x faster on output tokens, and is optimized for AI agents and coding tasks.

Google just released Gemini 3.5 Flash at Google I/O May, 2026. It is the first Gemini 3.5 model. The series combines frontier intelligence with action. Google calls it a major leap for intelligent agents. The Flash tier has historically been faster and cheaper. 3.5 Flash outperforms Gemini 3.1 Pro on challenging benchmarks. The previous premium tier has now been surpassed. What the Benchmarks Say Gemini 3.5 Flash scores 76.2% on Terminal-Bench 2.1. That benchmark tests coding performance.

Model ReleasesGoogle I/O 2026Agentic AI

Source details

Google just released Gemini 3.5 Flash at Google I/O May, 2026. It is the first Gemini 3.5 model. The series combines frontier intelligence with action. Google calls it a major leap for intelligent agents. The Flash tier has historically been faster and cheaper. 3.5 Flash outperforms Gemini 3.1 Pro on challenging benchmarks. The previous premium tier has now been surpassed. What the Benchmarks Say Gemini 3.5 Flash scores 76.2% on Terminal-Bench 2.1. That benchmark tests coding performance.

Model ReleasesGoogle I/O 2026Agentic AI

News Latent.Space May 20

[AINews] Google I/O 2026: Gemini 3.5 Flash, Omni (NanoBanana for Video), Spark (background agents), and Antigravity 2.0

By Unknown

90 score

AI Analysis

Continuing our coverage from yesterday, Comprehensive coverage of Google I/O 2026 announcements including Gemini 3.5 Flash GA, Omni (video/NanoBanana), Spark (background agents), and Antigravity 2.0. Google demonstrated industry-leading capabilities across voice, video, and image modalities.

The full keynote livestream was 2 hours, but as usual, The Verge has the best supercut down to 30 mins, which is very worthwhile to get a narrative sense:The mainline Gemini 3.5 Flash is GA today (very nice compared to some staged rollouts) and is sold as a decent step up even compared to 3.1 Pro, with 3.5 Pro coming next month. Perhaps more impressive were the Gemini Live (Voice) and Omni (Video) and Google Pics/Flow (Images/VFX/music) modalities, where Google demonstrated industry leading capa

Model ReleasesGoogle I/O 2026Multimodal AIAgentic AI

Source details

The full keynote livestream was 2 hours, but as usual, The Verge has the best supercut down to 30 mins, which is very worthwhile to get a narrative sense:The mainline Gemini 3.5 Flash is GA today (very nice compared to some staged rollouts) and is sold as a decent step up even compared to 3.1 Pro, with 3.5 Pro coming next month. Perhaps more impressive were the Gemini Live (Voice) and Omni (Video) and Google Pics/Flow (Images/VFX/music) modalities, where Google demonstrated industry leading capa

Model ReleasesGoogle I/O 2026Multimodal AIAgentic AI

News Ars Technica - All content May 20

Buckle up: Google is set to remake search with agentic AI in 2026

By Ryan Whitwam

82 score

AI Analysis

Continuing our coverage from yesterday, Google is fully committing to agentic AI in search, with AI Mode usage doubling since its launch. VP Liz Reid declared 'Google search is AI search' at I/O 2026, signaling the permanent transformation of the company's core product.

Last year marked the beginning of Google's explicit focus on AI search, and this year's I/O solidified that shift. As Google's search VP Liz Reid said during the keynote, "Google search is AI search." This change is well underway, and the very reasonable objections to this path will not dissuade the company. All the metrics that matter to Google say this is the right move. But at the end of the day, Google can get whatever outcome it wants because it's just that big and influential. Google start

Google I/O 2026Agentic AISearch

Source details

Last year marked the beginning of Google's explicit focus on AI search, and this year's I/O solidified that shift. As Google's search VP Liz Reid said during the keynote, "Google search is AI search." This change is well underway, and the very reasonable objections to this path will not dissuade the company. All the metrics that matter to Google say this is the right move. But at the end of the day, Google can get whatever outcome it wants because it's just that big and influential. Google start

Google I/O 2026Agentic AISearch

News MarkTechPost May 20

NVIDIA AI Releases Nemotron-Labs-Diffusion: A Tri-Mode Language Model with 6× Tokens Per Forward Over Qwen3-8B

By Asif Razzaq

80 score

AI Analysis

NVIDIA released Nemotron-Labs-Diffusion, a language model family unifying autoregressive, diffusion-based parallel, and self-speculation decoding in one architecture. It generates 6x tokens per forward pass compared to Qwen3-8B, available in 3B, 8B, and 14B parameter sizes.

NVIDIA researchers have released Nemotron-Labs-Diffusion, a language model family that unifies three decoding modes in one architecture. The model supports autoregressive (AR) decoding, diffusion-based parallel decoding, and self-speculation decoding. It is available in 3B, 8B, and 14B parameter sizes. The family includes base, instruct, and vision-language variants. Sequential Decoding Limits Throughput Standard autoregressive (AR) language models generate text one token at a time, left t

Model ReleasesInference OptimizationNVIDIAOpen Source

Source details

NVIDIA researchers have released Nemotron-Labs-Diffusion, a language model family that unifies three decoding modes in one architecture. The model supports autoregressive (AR) decoding, diffusion-based parallel decoding, and self-speculation decoding. It is available in 3B, 8B, and 14B parameter sizes. The family includes base, instruct, and vision-language variants. Sequential Decoding Limits Throughput Standard autoregressive (AR) language models generate text one token at a time, left t

Model ReleasesInference OptimizationNVIDIAOpen Source

News AI News May 20

Alibaba is designing AI chips around agents, and that changes what the race is actually about

By Dashveenjit Kaur

78 score

AI Analysis

Alibaba unveiled the Zhenwu M890 AI processor purpose-built for AI agents, delivering 3x performance over its predecessor. The chip is designed for long-context retention and multi-step agent coordination, paired with a multi-year silicon roadmap and new LLM.

Alibaba has unveiled a new AI processor built specifically for AI agents, pairing the chip announcement with a multi-year silicon roadmap and a new large language model, signalling that the company is building an integrated AI stack, not just filling a gap left by US export controls. The Zhenwu M890, developed by Alibaba’s semiconductor subsidiary T-Head, delivers three times the performance of its predecessor, the Zhenwu 810E, according to the company, as per Reuters report. But the pe

AI HardwareAgentic AIChina AIAlibaba

Source details

Alibaba has unveiled a new AI processor built specifically for AI agents, pairing the chip announcement with a multi-year silicon roadmap and a new large language model, signalling that the company is building an integrated AI stack, not just filling a gap left by US export controls. The Zhenwu M890, developed by Alibaba’s semiconductor subsidiary T-Head, delivers three times the performance of its predecessor, the Zhenwu 810E, according to the company, as per Reuters report. But the pe

AI HardwareAgentic AIChina AIAlibaba

News Ars Technica - All content May 20

China banned RTX 5090D V2 while Nvidia CEO Jensen Huang was visiting

By Zijing Wu in Hong Kong and Michael Acton in San Francisco

77 score

AI Analysis

Building on yesterday's News coverage, China banned Nvidia's RTX 5090D V2 gaming chip while CEO Jensen Huang was visiting China with Trump. The move supports domestic chipmakers like Huawei and Cambricon as China pushes back against degraded US export-controlled chips.

Beijing banned an Nvidia gaming chip while the company’s chief executive, Jensen Huang, was visiting China with Donald Trump last week, the latest salvo in the superpowers’ battle to dominate AI. The chip was added to a list of banned goods at China’s customs checkpoints last Friday, according to a copy of the document seen by the FT and two people with knowledge of the matter. The move highlights Beijing’s determination to keep out Nvidia’s chips, especially the degraded versions made to comply

GeopoliticsAI HardwareUS-China RelationsPolicy

Source details

Beijing banned an Nvidia gaming chip while the company’s chief executive, Jensen Huang, was visiting China with Donald Trump last week, the latest salvo in the superpowers’ battle to dominate AI. The chip was added to a list of banned goods at China’s customs checkpoints last Friday, according to a copy of the document seen by the FT and two people with knowledge of the matter. The move highlights Beijing’s determination to keep out Nvidia’s chips, especially the degraded versions made to comply

GeopoliticsAI HardwareUS-China RelationsPolicy

News Feed: Artificial Intelligence Latest May 20

SpaceX Is Spending $2.8 Billion to Buy Gas Turbines for Its AI Data Centers

By Paresh Dave

75 score

AI Analysis

SpaceX is spending $2.8 billion on gas turbines to power AI data centers for Elon Musk's xAI unit, amid complaints about carbon emissions. The investment supports xAI's ambitions to become a major cloud computing player.

The investment comes as Elon Musk’s AI unit faces complaints about the carbon-emitting units and looks to become a big player in cloud computing.

AI InfrastructureEnergyxAIInvestment

Source details

The investment comes as Elon Musk’s AI unit faces complaints about the carbon-emitting units and looks to become a big player in cloud computing.

AI InfrastructureEnergyxAIInvestment

News MarkTechPost May 20

Alibaba Qwen Team Introduces Qwen3.5-LiveTranslate-Flash: Real-Time Multimodal Interpretation Across 60 Languages at 2.8-Second Latency

By Asif Razzaq

73 score

AI Analysis

Alibaba's Qwen team released Qwen3.5-LiveTranslate-Flash achieving real-time multimodal interpretation across 60 languages at 2.8-second latency, with speech output in 29 languages. This represents a major expansion from the prior version's 18 input languages.

Simultaneous interpretation is one of the harder problems in applied AI. You’re asking a model to translate speech before the speaker has finished a sentence. Every extra second of delay breaks the illusion of real-time communication. Alibaba’s Qwen team has been chipping away at this with each release. Their latest model, Qwen3.5-LiveTranslate-Flash, brings that latency down to 2.8 seconds and expands input language coverage to 60 languages. qwen.ai/blog?id=qwen3.5-livet

Model ReleasesMultilingual AIAlibabaMultimodal AI

Source details

Simultaneous interpretation is one of the harder problems in applied AI. You’re asking a model to translate speech before the speaker has finished a sentence. Every extra second of delay breaks the illusion of real-time communication. Alibaba’s Qwen team has been chipping away at this with each release. Their latest model, Qwen3.5-LiveTranslate-Flash, brings that latency down to 2.8 seconds and expands input language coverage to 60 languages. qwen.ai/blog?id=qwen3.5-livet

Model ReleasesMultilingual AIAlibabaMultimodal AI

News aibusiness May 20

Alibaba Aims for Independence with New AI Chips, Model

By Esther Shittu

72 score

AI Analysis

Alibaba aims for AI chip independence with new processors and full-stack strategy to reduce reliance on Nvidia chips amid US export controls.

The chip highlights the vendor’s move toward a full-stack AI strategy and its efforts to wean itself off Nvidia AI chips.

AI HardwareChina AIAlibaba

Source details

The chip highlights the vendor’s move toward a full-stack AI strategy and its efforts to wean itself off Nvidia AI chips.

AI HardwareChina AIAlibaba

News Ars Technica - All content May 20

The Internet can't stop watching Figure AI's humanoid robots handling packages

By Jeremy Hsu

70 score

AI Analysis

Figure AI's week-long livestream of humanoid robots handling packages became a viral sensation, including a robot-vs-human competition. The demo showcased sustained autonomous package handling on a conveyor belt.

The robotics startup Figure AI has been livestreaming humanoid robots placing thousands of packages onto a conveyor belt for nearly a week—a spectacle that included a robot competing against a human intern at one point. The promotional robot demo has become a viral sensation among tech enthusiasts, spurring YouTube commenters to name the robots and the company to rapidly roll out related robot merchandise in response. Users on X have described the livestream in glowing terms, such as “the greate

RoboticsFigure AIHumanoid Robots

Source details

The robotics startup Figure AI has been livestreaming humanoid robots placing thousands of packages onto a conveyor belt for nearly a week—a spectacle that included a robot competing against a human intern at one point. The promotional robot demo has become a viral sensation among tech enthusiasts, spurring YouTube commenters to name the robots and the company to rapidly roll out related robot merchandise in response. Users on X have described the livestream in glowing terms, such as “the greate

RoboticsFigure AIHumanoid Robots

AI News Summary

Key Themes

All Ranked Signals

Google Introduces Gemini 3.5 Flash at I/O 2026: A Faster and Cheaper Model for AI Agents and Coding

[AINews] Google I/O 2026: Gemini 3.5 Flash, Omni (NanoBanana for Video), Spark (background agents), and Antigravity 2.0

Buckle up: Google is set to remake search with agentic AI in 2026

NVIDIA AI Releases Nemotron-Labs-Diffusion: A Tri-Mode Language Model with 6× Tokens Per Forward Over Qwen3-8B

Alibaba is designing AI chips around agents, and that changes what the race is actually about

China banned RTX 5090D V2 while Nvidia CEO Jensen Huang was visiting

SpaceX Is Spending $2.8 Billion to Buy Gas Turbines for Its AI Data Centers

Alibaba Qwen Team Introduces Qwen3.5-LiveTranslate-Flash: Real-Time Multimodal Interpretation Across 60 Languages at 2.8-Second Latency

Alibaba Aims for Independence with New AI Chips, Model

The Internet can't stop watching Figure AI's humanoid robots handling packages