Rewrite this intro so it sounds less corporate.
GPT-5ClaudeGeminiDeepSeek
Switch models mid-conversation
Hand a thread from one model to another with the full context intact. No copy-paste, no starting over.
Early access is open→
Every model.One place.
GPT, Claude, Gemini, DeepSeek, Llama and 40+ more — in one chat, with shared memory, side-by-side compare and a single subscription.
Free plan · No credit card · Switch models any time
GPT-5▾MemoryCompare
Write a launch post for our app. Punchy, under 280 characters.
GPT-5
Three directions — pick your vibe:
- Bold. Every AI model. One tab. Switch between GPT, Claude and Gemini mid-thought. Meet TunnelGPT.
- Witty. We sent 40 AI models into a tunnel. Only the best answer comes out the other side.
- Calm. One place for every model you rely on. Early access is open.
↑
Write a launch post for our app. Punchy, under 280 characters.
GPT-5
Three directions — pick your vibe:
- Bold. Every AI model. One tab. Switch between GPT, Claude and Gemini mid-thought. Meet TunnelGPT.
- Witty. We sent 40 AI models into a tunnel. Only the best answer comes out the other side.
- Calm. One place for every model you rely on. Early access is open.
Why does findUser always return undefined?
function findUser(users, id) {
users.forEach(u => {
if (u.id === id) return u;
});
}Claude Sonnet
The return exits the arrow callback, not findUser — and forEach ignores return values. Use find:
const findUser = (users, id) =>
users.find(u => u.id === id);It also stops at the first match instead of scanning the whole array.
Summarize this report in three bullets for the board.PDFQ3-market-report.pdf · 84 pages
Gemini Pro
- Growth: revenue up 23% quarter over quarter, led by enterprise contracts.
- Risk: SMB churn rose to 4.1% — pricing is the most cited reason.
- Ask: approve two hires for enterprise onboarding in Q4.
Want a one-slide version with a chart?
Poster: a neon tunnel at night, cinematic, 4:5.
Flux
Models
The whole frontier.
One subscription.
Stop juggling five subscriptions and five browser tabs. Every leading model for text, code, images, video and voice — available the day it’s ready.
GPT-5OpenAIClaude OpusAnthropicGemini ProGoogleDeepSeek V3DeepSeekLlama 4MetaMistral LargeMistralGrokxAIQwen 3AlibabaClaude SonnetAnthropicGemini FlashGoogleGPT-5OpenAIClaude OpusAnthropicGemini ProGoogleDeepSeek V3DeepSeekLlama 4MetaMistral LargeMistralGrokxAIQwen 3AlibabaClaude SonnetAnthropicGemini FlashGoogle
Gemini FlashGoogleDeepSeek R1DeepSeekKimi K2MoonshotSonarPerplexityFluxBlack Forest LabsStable DiffusionStabilityImagenGoogleVeoGoogleKlingKuaishouWhisperOpenAIGemini FlashGoogleDeepSeek R1DeepSeekKimi K2MoonshotSonarPerplexityFluxBlack Forest LabsStable DiffusionStabilityImagenGoogleVeoGoogleKlingKuaishouWhisperOpenAI
24
Text & reasoning
Frontier and open models for writing, analysis and long, careful thinking.
9
Code
Models tuned for programming, reviews and refactors in every major language.
8
Image
From photoreal to illustration — generate, edit and upscale.
6
Video & voice
Text-to-video, transcription and natural speech in 30+ languages.
Product
One chat that thinks
in every model.
“Debug this stack trace”“A haiku about Mondays”“Review this 300-page contract”“Prove this lemma”“Debug this stack trace”
Claude SonnetGPT-5Gemini ProDeepSeek R1Claude Sonnet
Auto picks for you
Not sure which model? Auto routes every message to the one that fits it best.
Compare side by side
One prompt, up to four models, answers streaming in parallel.
Prefers TypeScriptTeam of 6Tone: conciseLaunch: Jan 14
Memory that travels
Tell it once. Every model remembers your context, tone and projects.
PDFcontract-v4.pdfCSVrevenue-2026.csvPNGwireframe.png
Files, PDFs & vision
Drop in documents, spreadsheets and screenshots. Ask anything about them.
Generating… 62%
Create images & video
Flux, Stable Diffusion, Imagen, Veo, Kling — every studio-grade generator, same chat.
Private by design
Your chats never train any model. Encrypted in transit and at rest, deletable in one click.
Compare
One prompt.
Every perspective.
Send the same question to up to four models at once. Spot the best answer in seconds — or merge them into one.
YouExplain quantum tunneling to a ten-year-old.3 models
GPT-5
Imagine rolling a ball at a hill that’s too tall to climb. Normally it rolls back. But tiny particles like electrons are a bit like waves — and sometimes a little of the wave leaks through the hill and the particle pops out the other side.
Claude Opus★ Your pick
You can’t walk through a wall. Tiny particles sort of can — not by breaking it, but because they’re fuzzy. A particle isn’t in one exact spot; it’s spread out a little, and a tiny bit of that fuzz reaches past the wall. Every so often, the particle shows up there.
Gemini Pro
Think of a video-game ghost that sometimes glides through walls. Electrons aren’t ghosts, but at tiny sizes they follow different rules: there’s a small chance they appear on the other side of a barrier. The memory chip in your phone uses this every day!
Vote for a winner and Auto learns which models you trust for what.
How it works
From sign-up to
every model in a minute.
01
Create your account
One sign-up. No API keys, no five separate subscriptions, no setup.
02
Pick a model — or don’t
Choose from 40+ models, or let Auto route each message to the best one.
03
Switch, compare, keep going
Change models mid-thread. Your context and memory come along for the ride.
40+models in one chat
1subscription instead of five
4models per Compare run
0tabs to juggle
Pricing
Every model.
One simple price.
Free
For curious minds
$0$0forever
Start for free →
- Fast models: Gemini Flash, DeepSeek, Llama
- 50 messages a day
- 5 Compare runs a day
- 10 image generations a month
- 30-day chat history
Most popular
Pro
For people who live in AI
$14$11/ mo
Go Pro →
- All 40+ models — GPT-5, Claude Opus, Gemini Pro and more
- Unlimited Compare & Merge
- Memory across every model
- 300 images & 20 videos a month
- Files up to 100 MB, unlimited history
- Priority speed at peak hours
Team
Per seat, for teams of 3+
$29$23/ seat / mo
Start a team →
- Everything in Pro
- Shared workspaces & prompt library
- Admin console & usage analytics
- Centralised billing & invoices
- API access (beta)
Prices in USD. Yearly plans billed annually. Cancel any time.
FAQ
Questions,
answered.
Which models are available?
TunnelGPT includes 40+ models across text, code, image, video and audio — including the GPT, Claude, Gemini, DeepSeek, Llama, Mistral, Grok and Qwen families. We add new models as soon as they’re production-ready, usually within days of release.
Can I switch models in the middle of a conversation?
Yes. Pick a different model from the switcher at any point and it continues with the full conversation history. You can also @mention a model to get a single reply from it without switching.
Do you train on my conversations?
No. Your chats are never used to train models — ours or anyone else’s. Models receive only what they need to answer your message.
What happens when I hit the Free plan limit?
You can keep chatting with fast models, or upgrade to Pro for the full lineup. Free limits reset every day at midnight UTC.
Is there an API?
API access is in private beta for Team workspaces. Once your workspace is live, request access from the admin console.
Why is there an early access queue?
We open workspaces in waves so every model stays fast for everyone. Create an account to reserve yours — we’ll email you the moment it’s ready.
Journal
Fresh from
the tunnel.
Early access to TunnelGPT is open
Starting today you can reserve a workspace in the AI superapp that puts every major model behind one chat. Here’s what’s in the first wave and how invites work.
Why one model is never enough
Every model has a personality — strengths, blind spots and habits. The people getting the most out of AI stopped picking a favourite a long time ago.
Introducing Compare: one prompt, every perspective
Send a single prompt to up to four models at once and read the answers side by side. Rolling out to beta members today.
Step into
the tunnel.
Create a free account and reserve your workspace in early access.
Get started — it’s free →