SendTech Times
العربية
AINews|May 24, 2026, 02:57 PM
AI SHIFT:

Gemini 3.5 Flash: Performance Boost Comes with Higher Costs

BySendTech Times AI & Enterprise DeskNewsroom-edited, source-reviewed coverage
Newsroom brief

Google's Gemini 3.5 Flash has shown a significant performance leap, but the cost of running it has skyrocketed by 5.5 times compared to its predecessor. Despite its advanced capabilities, the model's expenses for agent tasks are 75% higher than the Gemini 3.1 Pro. For readers, the financial implications of adopting such AI technologies are crucial as they balance performance with operational costs.

Verified against source materialEdited by SendTech Times AI & Enterprise Desk
Gemini 3.5 Flash: Performance Boost Comes with Higher Costs

5.5 times is the increase in operational costs for Google’s Gemini 3.5 Flash, a model that was unveiled at I/O 2026.

While this new version has improved its performance, the financial implications are significant, raising questions about its practicality for users.

The Cost of Innovation

The latest report from The Decoder highlights that Gemini 3.5 Flash, while outperforming its predecessor, comes with a hefty price tag.

Users will find that executing the same tasks now requires 5.5 times more investment.

This increase is particularly pronounced in agent tasks, where costs exceed those of the upper-tier Gemini 3.1 Pro by 75%.

Performance Metrics

On the performance front, Gemini 3.5 Flash has outpaced all reported models in coding and agent-centric evaluations.

It has surpassed competitors like Anthropic's Claude 4.7 and OpenAI's GPT-5.5, showcasing its capabilities in specific areas.

However, it does lag in broader knowledge tasks, such as processing long texts and complex exams.

Benchmarking Against Competitors

Google's benchmarks reveal that while Gemini 3.5 Flash excels in coding and agent tasks, it struggles with tasks requiring extensive knowledge.

This duality raises questions about the model's versatility and its suitability for various applications.

The pricing strategy reflects a middle ground, aiming to balance performance with affordability.

Implications for the Gulf Region

For businesses in the Gulf and MEA regions, the cost dynamics of AI technologies like Gemini 3.5 Flash are critical.

As companies seek advanced AI solutions, understanding the balance between performance and cost will shape their investment strategies.

The financial implications of adopting such technologies will be a key consideration for decision-makers in the region.

Next Steps in AI Development

As Google continues to refine its AI offerings, the focus will be on whether the performance gains justify the increased operational costs.

Monitoring how these changes affect user adoption and market competition will be essential for stakeholders in the tech industry.

Share this article
inXf

Related articles

More
Tencent Takes WorkBuddy AI Agent Global In Enterprise Productivity Push
AI

Tencent Takes WorkBuddy AI Agent Global In Enterprise Productivity Push

Tencent Cloud launched WorkBuddy for overseas users after an earlier China rollout. The agent can run tasks through messaging apps and connect with GitHub, Jira, Google Drive, Gmail, Notion, and Slack. Miora and TokenHub show Tencent building a wider enterprise AI stack around agents, creative work, and model access.

Microsoft Finds Cheaper AI Model Rates Can Still Raise Agent Costs
AI

Microsoft Finds Cheaper AI Model Rates Can Still Raise Agent Costs

Microsoft said lower token prices for Claude Sonnet 5 did not remove AI agent cost spikes when it compared Claude models inside GitHub Copilot.

MiniMax M3 turns long-context AI into an agent platform test
AI

MiniMax M3 turns long-context AI into an agent platform test

MiniMax launched M3 on June 1, 2026, combining long-context, agentic, coding and native multimodal capabilities in one model line. The API supports up to 1 million tokens of context, with a guaranteed minimum of 512K tokens, and includes M3 and M3-highspeed versions. MiniMax plans to open-source M3 on HuggingFace and GitHub, while early pricing offers a 50% discount for the first seven days.

Om AI Bets on Edge Multimodal Models as China AI Startups Move Toward Deployment
AI

Om AI Bets on Edge Multimodal Models as China AI Startups Move Toward Deployment

Om AI Technology is focusing on compact edge-side multimodal vision models for PCs, cameras, robots and other devices rather than very large cloud models. At BEYOND Expo 2026, the company showed OttoBox AI Studio, a local-AI content tool for video analysis, asset matching, script generation and fast production. The next test is whether its VLX edge multimodal model can improve video understanding and decision-making while keeping operating costs lower.

X-Square WALL-WM Points Robotics AI Toward Event-Level Planning
AI

X-Square WALL-WM Points Robotics AI Toward Event-Level Planning

X-Square Robot released WALL-WM, an embodied AI world model that predicts semantic events rather than fixed motion frames. The company says the approach helps robots focus on task objectives such as grasping an object instead of memorizing pixel-level movement sequences. Reported benchmarks show stronger motion quality, semantic consistency, physical plausibility and task completion than several comparison models.

Artificial Intelligence News Updates: Latest News About Google AI, OpenAI, ChatGPT, Gemini, Lamda and More
AI

Artificial Intelligence News Updates: Latest News About Google AI, OpenAI, ChatGPT, Gemini, Lamda and More

Pope Leo XIV is set to release a manifesto on artificial intelligence titled 'Magnificent Humanity'. The document aims to tackle ethical and social challenges posed by rapid AI development. This release follows years of study by the Church on AI-related technologies.

Keep Reading

More Stories

Latest
Rapidus Targets 2027 2nm Fab Ramp Without Volume CustomerChips & SemiconductorsJul 9, 2026Rapidus Targets 2027 2nm Fab Ramp Without Volume CustomerRapidus has EUV equipment running at its Chitose fab and plans 2nm mass production in 2027, but Tom's Hardware reported that the Japanese chipmaker still lacks a committed high-volume customer.Sony Wins OCC Conditional Trust Charter For Stablecoin UnitFintech & Digital PaymentsJul 9, 2026Sony Wins OCC Conditional Trust Charter For Stablecoin UnitSony disclosed that the OCC has conditionally approved a U.S. national trust bank for Connectia Trust, a Sony Financial Group subsidiary that would issue and manage dollar-denominated stablecoins. Sony said the unit will be capitalised at $40 million, while trade groups and consumer advocates objected to the charter path.Mastercard Says 77% Of Saudi SMEs Accept Online PaymentsFintech & Digital PaymentsJul 8, 2026Mastercard Says 77% Of Saudi SMEs Accept Online PaymentsMastercard research says 75% of Saudi SME leaders remain confident despite a fall from a 92% benchmark, while 77% accept online payments and the Saudi sample size is not disclosed.Naver And KAI Sign Defence AI MOU Without Weapons-Control TimelineAIJul 8, 2026Naver And KAI Sign Defence AI MOU Without Weapons-Control TimelineNaver, Naver Cloud and KAI signed a defence AI MOU for sovereign models and physical AI combat platforms, but the companies did not disclose budget, deployment timing, test results or weapons-control approvals.Elisa Claims AI Automation Cut Network Incidents By More Than 80%Telco & ConnectivityJul 8, 2026Elisa Claims AI Automation Cut Network Incidents By More Than 80%RCR Wireless News reported that Elisa says AI-powered automation has reduced customer-impacting network incidents by more than 80% and pushed preventive actions to 99%, while the operator still has not deployed Open RAN.AI Times Says Infrastructure Captures 82% Of Generative AI ValueCloud & Data CentersJul 8, 2026AI Times Says Infrastructure Captures 82% Of Generative AI ValueAI Times Korea cited Exponential View data showing 82% of measured AI-economy value going to cloud, GPU and inference infrastructure in Q1 2026, while foundation-model companies accounted for 11% and applications 7%.Klarna Seeks U.S. Industrial Bank Charter As Partner-Bank Model Faces ReviewFintech & Digital PaymentsJul 8, 2026Klarna Seeks U.S. Industrial Bank Charter As Partner-Bank Model Faces ReviewBanking Dive reported that Klarna applied for a U.S. industrial bank charter after the company said it had extended $91.3 billion in credit to U.S. consumers; the application still needs FDIC and Utah approval.SambaNova Raises $1 Billion For AI Inference Hardware Without IPO TimingChips & SemiconductorsJul 8, 2026SambaNova Raises $1 Billion For AI Inference Hardware Without IPO TimingTechCrunch reported that SambaNova Systems raised $1 billion at an $11 billion valuation, named JPMorganChase as an inference-infrastructure partner and did not give IPO timing or SN50 customer shipment dates beyond the second half of 2026.FuriosaAI Starts RNGD Accelerator Deployment At Equinix Lisbon DatacenterChips & SemiconductorsJul 8, 2026FuriosaAI Starts RNGD Accelerator Deployment At Equinix Lisbon DatacenterFuriosaAI has begun deploying its RNGD AI accelerators at Equinix’s LS2 datacenter in Lisbon as the South Korean chip startup looks for European sovereign AI demand. The company describes 48 GB of HBM3, 1.5 TB/s memory bandwidth and 512 teraFLOPS of dense FP8 performance per card, but its Broadcom-linked next-generation accelerator still depends on HBM4 and HBM4e timing.Kraken Targets European Banking Licence In 10-Year Licence PushFintech & Digital PaymentsJul 8, 2026Kraken Targets European Banking Licence In 10-Year Licence PushKraken is pursuing a full European banking licence with Lithuania as the focus jurisdiction, according to a person familiar with the plans. CEO Arjun Sethi said Kraken’s parent company aims to acquire licences over the next 10 years, while Kraken declined to comment and Lithuania’s central bank said licensing processes are confidential.Meta Muse Image Uses Public Instagram Posts Unless Users Opt OutCapital & PolicyJul 8, 2026Meta Muse Image Uses Public Instagram Posts Unless Users Opt OutMeta has launched Muse Image with Instagram integration that allows public profile photos and videos to be used in AI image generations by default. Users can switch off sharing in Instagram settings, but Meta says existing AI images made with their content will not be deleted and users will not be notified when someone creates them.IBM Adds Rackmount z17 And LinuxONE 5 Systems Without Customer CountsChips & SemiconductorsJul 8, 2026IBM Adds Rackmount z17 And LinuxONE 5 Systems Without Customer CountsIBM is extending its z17 and fifth-generation LinuxONE families with single-frame and rackmount systems, including LinuxONE Express as an 18U entry point. The smaller hardware brings Telum II and Spyre AI capability to lower-scale deployments, but IBM did not disclose customer counts, pricing or shipment volumes for the new models.