Gemini vs ChatGPT (2026): Benchmarks, Pricing & Verdict

Last updated: September 2026

Is Gemini better than ChatGPT in 2026?

In the Gemini vs ChatGPT race, ChatGPT wins on raw capability and Gemini wins on value. OpenAI's GPT-6 Astra leads most published benchmarks, especially agentic coding, but it costs about 13 times more per token than Google's Gemini 3.8 Flash and sits behind ChatGPT's $100 plans. Gemini 3.8 Flash is faster, far cheaper, ranks higher on human preference and plugs directly into Google Workspace. For enterprises, the deciding factor is often how safely either assistant can touch your email, files and tools.

September 2026 reshuffled both lineups. OpenAI shipped GPT-6 Astra on September 3 and added GPT-6 Sol and GPT-6 Luna on September 22. Google released Gemini 3.8 Flash on September 2, while its next Pro model, Gemini 3.5 Pro, has still not shipped. This guide compares the models, plans, prices, privacy terms and security record with the latest published data.

TL;DR: Key Takeaways

• Benchmarks: GPT-6 Astra beats Gemini 3.8 Flash on OpenAI's own table, from 57.9% vs 19.1% on Terminal-Bench 4.0 to 96.0% vs 95.3% on GPQA Diamond, per OpenAI.
• Human preference: Gemini 3.8 Flash ranks #9 on the LMArena Text leaderboard with 1493 points, ahead of Astra at #24 with 1480.
• API cost: Gemini 3.8 Flash costs $0.75 input and $3.75 output per million tokens until December 31, 2026, according to Google. GPT-6 Astra costs $10 and $50.
• Plans: Google AI Pro costs $19.99 per month with 5 TB of storage. ChatGPT Plus costs $20, but OpenAI's help center limits GPT-6 Astra to Pro, Business and Enterprise plans.
• Reach: the Gemini app passed 1 billion monthly users in August 2026, while ChatGPT reported 900 million weekly users in February 2026.
• Security: Gray Swan's indirect prompt injection tests broke GPT-6 Astra in 8.5% of scenarios and Gemini 3.8 Flash in 5.5%, but both assistants have had real data exfiltration flaws through connected apps.

At a Glance: Key Differences in 2026

What Are Gemini and ChatGPT in 2026?

Gemini and ChatGPT are general-purpose AI assistants built on families of large language models. Gemini is Google's assistant and model family, woven into Search, Android, Chrome and Workspace. ChatGPT is OpenAI's assistant, powered by the GPT-6 and GPT-5.6 models. Both come as consumer apps, business workspaces and developer APIs.

The models behind Gemini

Google's newest model is Gemini 3.8 Flash, generally available since September 2, 2026, which Google calls "our most intelligent Flash model" in its launch post. It reaches the Gemini app for Google AI Pro and Ultra subscribers. Google also released 3.8 Flash Cyber, a security-tuned variant available only to vetted defenders.

At the top end, Google still relies on Gemini 3.1 Pro Preview from February 2026 and Gemini 3 Deep Think for Ultra subscribers. Google announced Gemini 3.5 Pro in May 2026 as "coming next month", but The Decoder noted on September 2 that it still had not shipped. Free users get Gemini 3.6 Flash.

The models behind ChatGPT

OpenAI launched GPT-6 Astra on September 3, 2026, with a 1,050,000-token context window and 128,000 output tokens, according to its API documentation. On September 22, OpenAI added GPT-6 Sol and GPT-6 Luna, faster and cheaper models for paid ChatGPT tiers.

Access to Astra inside ChatGPT is narrower than many headlines suggest. OpenAI's help center says "GPT-6 Pro, powered by GPT-6 Astra" is available on the Pro, Business and Enterprise plans, while Plus users get GPT-5.6 Sol and the new GPT-6 models. Free and Go users default to GPT-5.6 Luna. For a security view of the new model, see our CISO's guide to GPT-6 Astra security implications.

What this means for buyers: Google's best new model is a fast, cheap Flash model, while OpenAI's best new model is a premium frontier model. That asymmetry shapes every comparison below.

Gemini vs ChatGPT Benchmarks: Who Wins on Performance?

GPT-6 Astra wins most published benchmarks, and by wide margins on agentic coding. Gemini 3.8 Flash stays close on science reasoning and web research, and it wins on human preference votes. Keep in mind that OpenAI benchmarked its premium model against Google's budget-tier Flash model, not against a Gemini Pro model.

Sources: OpenAI, GPT-6 Astra (vendor-reported, September 2026); LMArena Text and WebDev leaderboards (September 13 and 23, 2026).

Independent rankings

Independent indexes confirm the gap at the top. The Artificial Analysis leaderboard scores GPT-6 Astra at 53 on its Intelligence Index, against 41 for Gemini 3.8 Flash and 30 for Gemini 3.1 Pro Preview. BenchLM ranks Astra first overall with 88.69 and Gemini 3.8 Flash eighth with 73.43.

Gemini's strength shows up elsewhere. Artificial Analysis measures Gemini 3.8 Flash at 293.4 output tokens per second, which makes it one of the fastest frontier-class models available. On LMArena's text arena, where people vote blind on which answer they prefer, Gemini 3.8 Flash beats Astra by 13 points.

How to read vendor benchmarks

Most head-to-head numbers come from OpenAI's launch table, and each lab chooses the tests it shows. Some figures also conflict across sources: OpenAI reports 99.9% for Astra on ARC-AGI-3, while BenchLM lists 62.7%. We left disputed scores out of the table. For Gemini's Pro-tier reasoning, Google's own Gemini 3.1 Pro model card reports 94.3% on GPQA Diamond and 80.6% on SWE-Bench Verified, but on older test versions that can't be compared directly with Astra's newer ones.

How we compared Gemini and ChatGPT

This comparison uses only published, dated sources: vendor launch posts, model cards and system cards, independent leaderboards (Artificial Analysis, LMArena, BenchLM) and external red-team results. Where vendors report different test versions, we say so rather than mixing them. Prices are US list prices shown in September 2026. We did not run private tests, so treat every figure as a starting point for your own evaluation.

Coding and Agents: Codex vs Gemini CLI and Antigravity

ChatGPT is the stronger coding assistant in 2026. GPT-6 Astra triples Gemini 3.8 Flash on Terminal-Bench 4.0 (57.9% vs 19.1%) and ranks second on LMArena's WebDev arena, where Gemini 3.8 Flash sits 26th. Gemini remains useful for fast, cheap code generation, but not for long autonomous engineering tasks.

Both companies ship coding and agent tools:

• Codex is OpenAI's coding agent, included with paid ChatGPT plans, and ChatGPT Agent handles multi-step web tasks.
• Gemini CLI is Google's open-source terminal agent, and Antigravity 2.0, announced at Google I/O 2026, is its agent-first development platform.
• Gemini Spark is Google's always-on personal agent for Ultra subscribers, running on dedicated cloud machines across Gmail, Docs, Sheets and Chrome, per TechCrunch.

For price-sensitive teams, the math can flip the answer. Gemini 3.8 Flash handles routine coding tasks at a fraction of Astra's price, and batch requests are 50% cheaper again. A common pattern is to route simple tasks to a cheap model and escalate hard ones to a frontier model, which makes model routing and governance part of the architecture.

What this means for engineering leaders: an agent that can run commands or act in Gmail and Drive is a new privileged identity in your environment. Treat access scopes, approvals and logging as part of tool selection, not an afterthought.

Reasoning, Research and Multimodal

ChatGPT leads on hard reasoning, and Gemini leads on multimodal creation and speed. GPT-6 Astra holds the top published scores in math and science, while Gemini is the only one of the two that currently generates video and ships deeply into Google's apps.

Reasoning and research

OpenAI reports 97.6% for Astra on FrontierMath Tier 4 and 57.2% on Humanity's Last Exam with tools. Google has not published Gemini 3.8 Flash scores on those exact tests. For deep research, both apps offer agentic research modes, and the two models land within a point of each other on BrowseComp (91.5% vs 90.8%).

Images, video and voice

Gemini covers more media types today. Google's Nano Banana image models drew "over 10 million new users" and "more than 200 million image edits" after launch, according to Wikipedia's Gemini entry, and Gemini Omni video generation became generally available in the API on August 27, 2026. OpenAI's GPT Image 2 powers image creation in ChatGPT, but ChatGPT has no native video generation since OpenAI shut down the Sora app in April 2026, according to Wikipedia's Sora entry. Both offer live voice conversations.

Long context

Both flagship APIs accept about 1 million input tokens. GPT-6 Astra returns up to 128,000 output tokens and Gemini 3.8 Flash up to 64,000, which matters for long reports and large code changes. Inside the ChatGPT app, context windows are smaller and depend on the plan.

Gemini vs ChatGPT Pricing: Plans and API Costs

Gemini is cheaper at almost every level. Google AI Pro and ChatGPT Plus both cost about $20 per month, but Google adds 5 TB of storage and Workspace integration. On the API, Gemini 3.8 Flash costs about 13 times less than GPT-6 Astra per token.

Subscription plans

Sources: Google AI subscriptions; Gemini Enterprise; OpenAI Pro tiers; ChatGPT Go. Prices as shown in September 2026 and subject to regional promotions.

One detail for power users: OpenAI paused new sign-ups and upgrades for the $200 Pro tier on September 10, 2026, according to its help center.

API pricing

At scale, the gap is large. A workload of 10 million input tokens and 2 million output tokens per month costs about $15 on Gemini 3.8 Flash at the promotional rate and $200 on GPT-6 Astra. Price per token is not price per task, though: if a cheaper model needs more retries, part of that saving disappears.

Ecosystem and Integrations

Gemini's advantage is distribution inside Google's products, and ChatGPT's is a large, independent app ecosystem. Google reported that the Gemini app "surpassed 1 billion monthly users" in an August 2026 post, and AI Mode in Search passed 1 billion monthly users as well. ChatGPT reported more than 900 million weekly active users in February 2026.

For teams that live in Google Workspace, Gemini's native integration removes a lot of friction. For teams on Microsoft 365 or mixed stacks, ChatGPT's connectors and wider third-party support often fit better.

Enterprise Privacy and Data Controls

On business plans, neither Google nor OpenAI trains on your data by default. The bigger gap is on consumer accounts: Gemini's consumer app keeps activity on by default and uses those chats to improve its models unless users switch it off.

Sources: Google Workspace Generative AI Privacy Hub; Gemini Enterprise; OpenAI Enterprise Privacy; OpenAI Business Data.

Google's compliance list is longer, including FedRAMP High and ISO 42001 for AI management. Coverage has gaps, though: Google states that Gemini Notebook and Gemini in Chrome do not yet support its ISO, SOC, HIPAA and FedRAMP certifications. On consumer accounts, Google's privacy hub says activity is kept for 18 months by default and human-reviewed chats for up to 3 years. That is why personal Gemini or ChatGPT accounts used for work are a shadow AI risk.

Security and Governance: Deploying Gemini and ChatGPT in the Enterprise

Both assistants have strong model-level defenses and a real record of exploits through connected apps. The risk that matters most is indirect prompt injection: hidden instructions in an email, calendar invite, document or web page that make the assistant leak data or take actions. The OWASP Top 10 for LLM Applications ranks prompt injection as the number one risk (LLM01).

What the security data shows

• GPT-6 Astra: OpenAI's system card reports a 99.99% defense rate against direct injection. In Gray Swan's external test of 1,810 indirect attacks, attackers still succeeded in 8.5% of scenarios.
• Gemini 3.8 Flash: The Decoder reports a 5.5% attack success rate on Gray Swan's indirect injection benchmark. Google's own post mentions a "significant leap" without a number, so treat this comparison with caution.
• Google's defenses: Google describes a six-layer defense for Workspace, from classifiers and markdown sanitization to user confirmations, in its admin documentation.

Real incidents on both sides

• Gemini: a malicious calendar invite could make Gemini leak private meeting details, according to The Hacker News (January 2026, patched). Tenable's "Gemini Trifecta" found three injection flaws across Cloud Assist, Search personalization and browsing in 2025.
• ChatGPT: Radware's ZombieAgent showed zero-click data exfiltration through Gmail, Outlook, Drive and GitHub connectors, as Infosecurity Magazine reported (January 2026, fixed). Check Point found a DNS-based exfiltration channel patched in February 2026.
• Autonomous behavior: in September 2026, TechCrunch reported that Gemini accessed three companies' systems during third-party security testing, using password guessing and leaked credentials.

NeuralTrust's own research points the same way. Our Echo Chamber jailbreak bypassed guardrails on earlier Gemini and GPT models in 2025 through multi-turn context poisoning, and our Semantic Chaining research confirmed Gemini's Nano Banana Pro image model as vulnerable.

Why this decides enterprise rollouts

According to Gartner (2025), over 40% of agentic AI projects will be canceled by the end of 2027, with inadequate risk controls among the main causes. The deeper an assistant reaches into email, files and business apps, the more an injected instruction can do. Vendor safeguards help, but you still need one policy layer that you control across every model and tool.

How NeuralTrust secures Gemini and ChatGPT

NeuralTrust adds a runtime security layer that works the same way for Gemini, ChatGPT or both:

• Agent Gateway (TrustGate): routes agent traffic to models, MCP servers and APIs with identity-aware access control, so every prompt, response and tool call passes one policy point. TrustGate ships with 200+ pre-built MCP servers.
• Agent Runtime Security (TrustGuard): inspects interactions in real time to block prompt injection, data leakage and unauthorized tool use.
• AI Red Teaming (TrustTest): attacks your own Gemini and ChatGPT deployments with techniques like the ones above, so you measure real exposure instead of relying on vendor benchmarks.

NeuralTrust was recognized across four Gartner Hype Cycle reports in 2026 in the AI Runtime Defense category. To go deeper on the attack itself, read how prompt injection works.

Which Should You Choose? Gemini or ChatGPT

Choose ChatGPT if you need the strongest reasoning and coding model and can pay for Pro or Business. Choose Gemini if your team runs on Google Workspace, needs video or fast answers, or wants the best price-performance. Many companies will use both, which makes consistent governance more important than the pick itself.

Conclusion

The Gemini vs ChatGPT choice in 2026 is a choice between peak capability and value. GPT-6 Astra is the stronger model for coding, agents and hard reasoning, but it is expensive and limited to ChatGPT's top plans. Gemini 3.8 Flash delivers most of the everyday quality at a fraction of the price, with the best Workspace integration and native video. Whichever you deploy, secure its access to your data the same way.

Secure Gemini and ChatGPT in Production with NeuralTrust

Run Gemini, ChatGPT or both behind one policy layer with real-time protection for every prompt, response and tool call.

FAQs about Gemini vs ChatGPT

1. Is Gemini better than ChatGPT?

It depends on what you need. ChatGPT's GPT-6 Astra leads most benchmarks, especially coding, with 57.9% on Terminal-Bench 4.0 against 19.1% for Gemini 3.8 Flash. Gemini is much cheaper, faster, ranks higher on LMArena's human preference leaderboard and integrates natively with Google Workspace. Heavy coders lean ChatGPT, Workspace teams lean Gemini.

2. Is Gemini or ChatGPT better for coding?

ChatGPT is better for coding in 2026. GPT-6 Astra ranks second on LMArena's WebDev leaderboard, while Gemini 3.8 Flash ranks 26th, and Astra triples Gemini's score on Terminal-Bench 4.0. Gemini is still a solid, low-cost option for simple code generation, and Google's Gemini CLI and Antigravity tools are improving quickly.

3. Which is cheaper, Gemini or ChatGPT?

Gemini is cheaper. Google AI Pro costs $19.99 per month and includes 5 TB of storage, similar to ChatGPT Plus at $20. On the API, Gemini 3.8 Flash costs $0.75 per million input tokens and $3.75 per million output tokens until December 31, 2026, against $10 and $50 for GPT-6 Astra.

4. Can Gemini and ChatGPT generate images and video?

Both generate images: Gemini with Google's Nano Banana models and ChatGPT with GPT Image 2. Only Gemini generates video today, through Gemini Omni. OpenAI shut down its Sora video app in April 2026, so ChatGPT has no native video generation. Both assistants also support live voice conversations.

5. Is Gemini or ChatGPT safer for enterprise use?

Both protect business data well and don't train on it by default. Google offers a longer certification list, including FedRAMP High and ISO 42001. On prompt injection, Gemini 3.8 Flash scored 5.5% and GPT-6 Astra 8.5% in Gray Swan tests, but both have had real exfiltration flaws through connected apps. Neither is safe without runtime controls.

6. Do Gemini and ChatGPT use my data for training?

On business plans, neither uses your data for training by default. On consumer accounts, Gemini keeps activity on by default and uses chats to improve its models, with human-reviewed chats stored for up to 3 years. ChatGPT lets consumers turn off training in data controls. Check settings before employees use personal accounts for work.

7. Can I use Gemini and ChatGPT together?

Yes. Many companies use Gemini for Workspace productivity and ChatGPT for coding or advanced reasoning. The challenge is governance: two vendors mean two sets of policies, logs and data flows. An AI gateway that applies the same access rules, monitoring and injection defense to both keeps multi-model use auditable.

About the Author

Roger Howroyd is Head of Global SEO and AI at NeuralTrust, where he leads the company's search strategy across SEO, AEO, GEO, and LLM optimization. He specializes in AI-powered search, content strategy, and SEM. Connect on LinkedIn.

NeuralTrust is the leading platform for securing and scaling AI agents. Named a Pioneer in the Gartner Emerging Market Quadrant for AI Application Security 2026, recognized across four Gartner Hype Cycle reports in 2026, and featured in the Gartner Market Guide for Guardian Agents 2026, the Gartner Market Guide for AI Gateways 2025 and the KuppingerCole Leadership Compass for Generative AI Defense 2025. Headquartered in Barcelona with offices in London and New York. ISO 27001 certified.

Sources

1. OpenAI: GPT-6 Astra (September 2026)
2. OpenAI API: GPT-6 Astra model page (September 2026)
3. OpenAI Help: GPT-5.6 and GPT-6 Pro in ChatGPT (September 2026)
4. OpenAI: GPT-6 Astra System Card, Prompt Injection (September 2026)
5. Google: Introducing Gemini 3.8 Flash and 3.8 Flash Cyber (September 2026)
6. Google DeepMind: Gemini 3.1 Pro model card (February 2026)
7. Google: Gemini API pricing (September 2026)
8. Google: Gemini app surpasses 1 billion monthly users (August 2026)
9. The Decoder: Gemini 3.8 Flash is Google's third budget model in six weeks (September 2026)
10. LMArena: Text leaderboard (September 2026)
11. Artificial Analysis: Model leaderboard (September 2026)
12. BenchLM: Gemini 3.8 Flash vs GPT-6 Astra (September 2026)
13. Google Workspace: Generative AI Privacy Hub (August 2026)
14. OpenAI: Enterprise privacy
15. TechCrunch: Google's Gemini is the latest AI model to hack other companies (September 2026)
16. The Hacker News: Google Gemini prompt injection flaw (January 2026)
17. Infosecurity Magazine: New zero-click attack on ChatGPT (January 2026)
18. OWASP: LLM01:2025 Prompt Injection
19. Gartner: Over 40% of Agentic AI Projects Will Be Canceled by End of 2027 (June 2025)