Routing service that finds the cheapest reliable model for every AI request you send. 6gnvghh8LKoM59p1WZSuTgYmdJrnZnhU7BzCcEaEASY
Joined June 2026
- Tweets633
- Following68
- Followers723
- Likes283
Pinned Tweet
$SWCH is live.
kickstart.easya.io/token/6gn…
Switchboard is live. One OpenAI-compatible endpoint that auctions every request to the cheapest reliable model.
The savings are real - 99.7% on easy requests routed away from GPT-4o. The token ledger is on-chain next.
Point your SDK at the endpoint. Add one header. The market runs from there.
switchesboard.com/
Your dashboard is live. Every account has one now.
Your requests, your savings vs GPT-4o, solana:6gnvghh8LKoM59p1WZSuTgYmdJrnZnhU7BzCcEaEASY burned from your fees, all in real time with the receipts on-chain. Plus badges to unlock, daily streaks, and levels to climb from Operator on up. Burn your first 1,000 solana:6gnvghh8LKoM59p1WZSuTgYmdJrnZnhU7BzCcEaEASY and Furnace I is yours.
One tap turns your savings into a card you can post.
Go see what you’ve been saving, then drop your card below.
For two years the story was one line. Inference keeps getting cheaper. Tokens fell by thousands of times and nobody expected that to turn.
Look at the hardware under those tokens though. H100 rental is up about 40% in six months. On-demand capacity is nearly sold out. The wait for a cluster runs 12 to 18 months. What’s eating it is inference and agents, firing far more calls per task than anyone planned for.
Cheap tokens, scarce compute. Send a request to the wrong model and you pay twice. Once at the model’s price. Again in capacity that half the market is queuing for.
Having the strongest model isn’t the edge anymore. Not wasting scarce compute is. Each request on the cheapest model that can handle it, nothing frontier-priced that didn’t need to be.
That’s the whole reason the auction exists.
Numbers from the exchange, live, not a screenshot.
7.6K requests routed
98.6% average saved vs GPT-4o
4,369 solana:6gnvghh8LKoM59p1WZSuTgYmdJrnZnhU7BzCcEaEASY burned from real fees
1345x price spread we route across
Every figure here comes straight from the exchange in real time, not typed into a slide. The savings hold up because the receipts sit on-chain and anyone can check them. That part matters to us more than any number.
And we’ll be honest, this is the early innings. 7.6K requests is a start, not a finish. The auction is live, the savings are real, the token burns on real usage. Now we go make all of it bigger.
The market keeps moving. We’re just getting warmed up.
Our docs are live!
Everything about how Switchboard works under the hood is now in one place. How the auction prices each request, how the cheapest capable model wins, what the on-chain receipt proves, and how to plug in with one line of code.
No fluff, just how it works.
Choose wisely.
Here’s the paradox nobody priced in from b2b side of things... Inference just hit its lowest level of the year, around $1.16 per million tokens, down from $2.04 in the spring. And yet enterprise AI bills are climbing, not falling. Median spend on LLMs grew more than 7x year over year.
Cheaper tokens didn’t shrink the bill. They removed the barrier to burning far more of them. And agents poured fuel on it. One agent task isn’t a single call, it’s dozens, as it reads, plans, calls tools, checks itself, corrects. 92% of companies rolling out agents are already over budget. Not “will be one day.” Already.
The thing is, none of that spend is the problem. The problem is paying frontier prices for work a model many times cheaper would have handled, on every one of those dozens of steps.
That’s what we built @SwitchesBoard for. Every request goes to an auction, and the cheapest model that can do the job wins it. The market keeps getting cheaper. We make sure that lands on your bill instead of passing you by. So it's time to choose the path.
We updated how conversion works.
Before, the token you deposited to pay with was converted to $SWCH the moment it landed. Now it stays in its original token until you make a request.
Conversion to $SWCH happens when you actually pay for inference, at the rate at that moment. You hold your own token right up until you spend it.
More control over your funds, fewer conversions you never asked for.
.@token2049 Singapore just wrapped, and one theme was everywhere: AI agents and how they pay for things!
@beincrypto launched its State of AI Agent Payments 2026 report right at the event, and the whole week was stacked with AI, DePIN and compute panels.
The signal is clear: the market now treats agent infrastructure and inference economics as a core narrative, not a side bet. That's the exact ground we're building on, where every agent request gets auctioned to the cheapest capable model instead of overpaying by default.
The conversation is catching up to what we've been shipping.
Switch Board retweeted
@SwitchesBoard is actively helping Inferra get the perps ready for launch. They’ve shared their experience with us and assigned a developer to help review and finalize everything.
Switchboard now accepts solana:9ufM9TJd1UEmi9awnGfxCkCHAgQ3JZ5Sw6YxeSeEASY as payment.
@InferraTrade is the exchange for GPU compute - GPU hours and reserved capacity, priced and traded like commodities. Now you can pay for inference on Switchboard directly in solana:9ufM9TJd1UEmi9awnGfxCkCHAgQ3JZ5Sw6YxeSeEASY: connect your wallet, route your requests, no swapping.
Compute market meets inference market 🤝
We locked ~10% of solana:6gnvghh8LKoM59p1WZSuTgYmdJrnZnhU7BzCcEaEASY supply for 3 months again through @streamflow_fi
It's on-chain and anyone can verify it. No unlock tricks, no quiet moves.
We've said from the start that this is a long-term build. This is that commitment, made in something you can check rather than take on faith
Switchboard now accepts $SQUIRE as payment.
$SQUIRE is the token powering @UsePod - an agent-first AI inference marketplace where spare GPU and API capacity gets sold and bought per request, settled in USDC on Solana via x402. Agent-first software, built by agents, for agents.
Now their token works directly on Switchboard: deposit $SQUIRE, pay for inference on our exchange, skip the swap entirely. Two inference marketplaces, one less point of friction for the agents routing through both.
@clawpumptech powers agents. $SQUIRE powers their compute. And now $SQUIRE fuels inference on Switchboard too.
Switchboard now accepts $MTR as payment.
Metera is the trust layer for agent data - it picks the source, verifies the answer, and tells your agent how much to trust it. Now you can pay for inference on Switchboard with $MTR: connect your wallet, route your requests, skip the swap.
@metera_xyz 🤝
We've added $FATCAT as a payment option on Switchboard.
FatCat is a non-custodial Solana trading app built for messaging apps - trade swaps, perps and DCA with an AI assistant, without ever leaving the chat or giving up your keys. Now their community token works directly on our platform: deposit $FATCAT, pay for inference, no swapping needed
@FatCatTrading 🤝
We've added $PYTO as a payment option on Switchboard.
@PytoAI is building the infrastructure layer for AI training - labeled data done right. We're building the inference exchange. Now their token works directly on our platform: connect your wallet, deposit $PYTO and skip the swap entirely.
Two EasyA projects, one less point of friction. Building the ecosystem for real.
We've integrated $CVT as a payment option on Switchboard.
@LaunchCovenant is building the governance layer for AI agents. We're building the exchange those agents route inference through. Now their token works directly on our platform - deposit $CVT, we do the rest.
The agent economy works better when the pieces connect. This is one of them.
