ArcRouter

The onchain inference router for AI agents on @arc.

CA: 0x5681e38CBFF1d5c4461294c6CF3970AA0761D2EB

[ArcRouter @useArcrouter]

https://t.co/pkAURhWEzR

[Watcher.Guru @WatcherGuru]

JUST IN: BlackRock says AI agents will drive major demand for crypto stablecoins and blockchain payments.

[ArcRouter @useArcrouter]

@WatcherGuru We knew what AI agents needed, so we built it.
Inference, pay-per-use, settled in USDC on @arc.

https://t.co/Kea8Q8JQ1I

[ArcRouter @useArcrouter]

https://t.co/pkAURhWEzR

[ArcRouter @useArcrouter]

@arc Built so you can build.
https://t.co/Kea8Q8JQ1I

[ArcRouter @useArcrouter]

https://t.co/pkAURhWEzR

[ArcRouter @useArcrouter]

Claude Opus 5.5 is live on ArcRouter.

The first model on Claude 5.5 family. It performs at the level of Claude Fable 5.1 for most tasks, and costs 40% less to run than Opus 5.

No signup, no card. Pay per call in USDC on @arc, or let your agent pay the 402 itself. At half price.

[ArcRouter @useArcrouter]

@Arguspad @arc gArgus

[ArcRouter @useArcrouter]

Most agent tokens aren't the clever ones.

They're classification, routing, retries, summaries: the same small job thousands of times a day. That work shouldn't cost frontier prices.

glm-5.3-flash from @Zai_org is on ArcRouter for exactly that: $0.15 in / $0.50 out per 1M tokens, paid in USDC on @arc.

model="glm-5.3-flash"

[ArcRouter @useArcrouter]

1M tokens of deepseek-v4.1-flash from @deepseek_ai on ArcRouter: $0.30 in, $1.20 out.

That's roughly 2,000 agent calls for the price of a coffee, settled per request in USDC on @arc, and it reads images now.

Cheap enough to be the default. Fast enough to be the loop.

model="deepseek-v4.1-flash"

[ArcRouter @useArcrouter]

@ArcHubHB @arclightfi @SugraArgusArc @focidotfamily @ArcBullX @Solonlabs1 @Buildogmeme @TenOnArc @ArctideApp @obrainarc The first onchain inference router for AI agents on @arc.

https://t.co/Kea8Q8JQ1I

[ArcRouter @useArcrouter]

https://t.co/pkAURhWEzR

[ArcRouter @useArcrouter]

We’re just getting started.

[Arc Hub @ArcHubHB]

Top gainers on Arc last 24h

What are you bullish on?👇

$ROUTER @useArcrouter
$ARCLIGHT @arclightfi
$SUGRA @SugraArgusArc
$FOCI @focidotfamily
$BULL @ArcBullX
$ASTOCK @Solonlabs1
$BUILDOG @Buildogmeme
$TEN @TenOnArc
$TIDE @ArctideApp
$oBrain @obrainarc

[ArcRouter @useArcrouter]

AI agents think through a model. Everything you tell them lands in its prompt, and whoever runs the GPU can read it.

With ArcRouter, it's sealed. The model runs in a GPU enclave the host can't see into, with a signed receipt on every answer.

Private onchain inference for AI agents on @arc, live on ArcRouter.

[Kalshi @Kalshi]

JUST IN: Circle says USDC accounts for 98.8% of agent-driven transaction volume

[ArcRouter @useArcrouter]

USDC keeps expanding as infrastructure for internet-native commerce.

The next step is machines spending it.

On ArcRouter, AI agents can access inference, pay only for what they use, and settle every request in USDC on @arc through x402.

Intelligence becomes pay-per-use.
USDC becomes machine money.

[Circle @circle]

Circle 🤝 @binance

Circle and Binance are continuing to build together through a new five-year commercial agreement to expand USDC access across emerging markets.

Binance has also made a $100M strategic investment in Circle.

https://t.co/I0CIUBUoCZ

[ArcRouter @useArcrouter]

@circle @binance Bullish for USDC

[ArcRouter @useArcrouter]

Six models you can now run privately on ArcRouter. Same API, just prefix the model with private/.

Sealed in GPU enclaves the host can't see into, with a signed receipt on every call. From $0.15 per 1M tokens, paid in USDC on @arc.

[ArcRouter @useArcrouter]

Private inference is live on ArcRouter.

Add "private/" to the model name and your call runs inside a GPU enclave the host can't see into.

Never cached. Prompts never logged. Signed receipt on every answer.

Six models from $0.15/1M, paid in USDC on @arc.

[ArcRouter @useArcrouter]

What it protects: your prompt from the machine's operator, and from anyone hoping we kept a copy. We never cache or log it.

What it doesn't: ArcRouter still routes the request in memory, like any gateway.

Docs: https://t.co/U9EGUA3CSk

[ArcRouter @useArcrouter]

Private inference is live on ArcRouter.

Add "private/" to the model name and your call runs inside a GPU enclave the host can't see into.

Never cached. Prompts never logged. Signed receipt on every answer.

Six models from $0.15/1M, paid in USDC on @arc.

[ArcRouter @useArcrouter]

How it works:

→ call any private/* model, same API
→ it runs in an Intel TDX + NVIDIA confidential GPU enclave, via @PhalaNetwork
→ the response carries a receipt

Check the receipt's chat_id matches your completion, and verify its signature against Phala's public attestation.

[ArcRouter @useArcrouter]

@fashiaka 🫡

[ArcRouter @useArcrouter]

Your prompts, sealed.

Private inference is coming to ArcRouter. Models running inside GPU enclaves the host can't see into, with a signed receipt on every answer.

Same endpoint. Same USDC on Arc.

Soon.

[ArcRouter @useArcrouter]

The machines chose USDC.

Now they need things to spend it on.

ArcRouter turns AI inference into something agents can buy per request with USDC via x402 on @arc.

No subscription. No account. Just machines paying for intelligence as they need it.

Agentic commerce is becoming real infrastructure.

[CoinMarketCap @CoinMarketCap]

LATEST: 🤖 Circle CEO Jeremy Allaire says "the machines have already chosen their money," pointing to agent-to-agent transactions overwhelmingly being settled in USDC.