Key Takeaways
- • NEAR AI reported on Oct. 21, 2026 that more than 1 million NEAR tokens are staked with its service, where staking rewards pay for AI inference while users retain their underlying tokens.
- • Inference credits accrue each second from staking yield, valued at $1 of platform usage per credit, and unused balances carry forward; unstaking stops new credits.
- • Credit budgets vary with the amount staked, NEAR's price, and staking APY, and are described by NEAR AI as illustrative rather than fixed.
- • Agent hosting uses a separate formula: 500 NEAR staked generates $5 in monthly credits, with a 50 NEAR minimum to activate a first agent, per the company's July 30 product explanation.
- • NEAR AI's pay-as-you-go catalog lists GLM 5.3 Flash at $0.08 per million input and $0.50 per million output tokens, and Qwen 3.8 27B at $0.04 and $1.35, with no minimum spend or subscription on that plan.
Summary
NEAR AI said on Oct. 21, 2026, in an X post that the amount of NEAR staked with its service had passed 1 million tokens, a figure NEAR Protocol repeated minutes later. The service lets users direct staking rewards toward AI inference while the deposited tokens remain in their ownership. "Stake yours to pay for your AI. The NEAR stays yours," NEAR AI wrote, according to the post as quoted by crypto.news.
The mechanism was introduced on July 30, when the protocol launched staking-based AI payments covering 43 models from providers including OpenAI, Anthropic and Google, alongside confidential inference and hosted agents. According to NEAR AI's inference documentation, staking yield is converted into computing credits. Credits accumulate continuously, unused balances carry forward, and unstaking halts new credit accrual. Each credit is valued at one US dollar of platform usage. Users manage changes through signed wallet transactions.
The reward rate is variable. NEAR AI says network rewards are divided among stakers, so the share of supply participating in staking affects yield. Typical validator commissions of 1% to 10% are deducted before rewards reach users. NEAR AI provides a calculator for translating a monthly inference allowance into required stake, though the company describes its figures as illustrative and directs users to the cloud console for current model prices.
Separately, NEAR AI publishes a pay-as-you-go price list, updated Oct. 21, with no minimum spend, subscription or seat license and no platform fee above the listed model rate. Using published rates, a workload of 1 million input and 1 million output tokens would cost $0.58 on GLM 5.3 Flash or $1.39 on Qwen 3.8 27B. The company states that secure execution and hardware-signed attestations are included in listed rates, and that its enterprise tier offers reserved GPU capacity, private models, volume pricing and service-level agreements.
For hosted agents, NEAR AI's July 30 explanation sets a different rule: the NEAR staked is divided by 100 to determine a monthly dollar credit budget, with 500 NEAR generating $5 per month. The company describes this as current policy rather than a permanent conversion rate. Staking also determines how many IronClaw agents can run in parallel.
The mechanism was introduced on July 30, when the protocol launched staking-based AI payments covering 43 models from providers including OpenAI, Anthropic and Google, alongside confidential inference and hosted agents. According to NEAR AI's inference documentation, staking yield is converted into computing credits. Credits accumulate continuously, unused balances carry forward, and unstaking halts new credit accrual. Each credit is valued at one US dollar of platform usage. Users manage changes through signed wallet transactions.
The reward rate is variable. NEAR AI says network rewards are divided among stakers, so the share of supply participating in staking affects yield. Typical validator commissions of 1% to 10% are deducted before rewards reach users. NEAR AI provides a calculator for translating a monthly inference allowance into required stake, though the company describes its figures as illustrative and directs users to the cloud console for current model prices.
Separately, NEAR AI publishes a pay-as-you-go price list, updated Oct. 21, with no minimum spend, subscription or seat license and no platform fee above the listed model rate. Using published rates, a workload of 1 million input and 1 million output tokens would cost $0.58 on GLM 5.3 Flash or $1.39 on Qwen 3.8 27B. The company states that secure execution and hardware-signed attestations are included in listed rates, and that its enterprise tier offers reserved GPU capacity, private models, volume pricing and service-level agreements.
For hosted agents, NEAR AI's July 30 explanation sets a different rule: the NEAR staked is divided by 100 to determine a monthly dollar credit budget, with 500 NEAR generating $5 per month. The company describes this as current policy rather than a permanent conversion rate. Staking also determines how many IronClaw agents can run in parallel.
Go deeper than the headline
Research every firm in this story, track the ones that matter and get the daily briefing — free for 7 days.