OYM Live Data · Morpheus project review

Live Inference Activity

Daily on-chain activity on the Morpheus Inference Contract on Base, reconstructed from every SessionClosed event since the December 2025 redeploy. Plus the access side: MOR locked into the Morpheus Marketplace API builder subnet for daily IPS credits at app.mor.org. Refreshed daily.

Glossary

  • Session. One paid run of an AI prompt through a Morpheus provider. Opens when the user (or the API gateway acting on their behalf) picks a provider and locks collateral; closes when the model finishes generating. Each session is a record on Base. Why care: a session is the unit of real demand on the network.
  • Provider. Anyone running a GPU node that serves inference. Each provider has its own Base address. Counting distinct addresses shows how many independent operators are actually closing work. Why care: Morpheus's freedom score depends on this not being one or two providers in a trench coat.
  • MOR earned. What a provider is owed when a session closes, calculated strictly on-chain as (min(closedAt, endsAt) − openedAt) × pricePerSecond. The "% claimed" badge tracks how much of that has actually been pulled to the provider's wallet. Why care: bottom-line economic activity. What was paid for inference and what a provider got for delivering it.
  • Tokens processed. The input and output token counts the provider signs into the session's closeout receipt. We read the attested totals straight from the receipt, the same on-chain source tech.mor.org reports from. Why care: the closest measure of real work done, but the provider self-reports it; the chain only enforces that they signed it.

The Morpheus team's tech.mor.org publishes a throughput dashboard from the same on-chain SessionClosed receipts. OYM indexes that source independently and adds the framing the raw counts skip: which figures are contract-enforced (MOR earned) versus provider-attested (tokens), and how much of the demand is emissions-subsidised rather than user-paid.

Live Inference Activity

Daily activity on the Inference Contract. Top row is the MOR providers earned for sessions that closed that day (capped at session endsAt). Bottom row is provider-attested tokens served (input + output counts from the signed closeout receipt).

MOR earned
15.43K
last 60 days · 51% claimed
Tokens (attested)
8.47B
100% of sessions
Sessions
85,663
12 unique providers all-time

MOR earned per day

Trustless · on-chain math
1.3K6260

Tokens served per day

Provider-attested · signed receipts
410.8M205.4M0
28 Apr 10 May 22 May 2 Jun 15 Jun 27 Jun
Mined from SessionClosed events on the LumerinDiamond Inference Contract at 0x6aBE…030a via Alchemy and the public Base RPC. MOR earned is on-chain math: (min(closedAt, endsAt) − openedAt) × bid.pricePerSecond, the entitlement for work delivered. The "51% claimed" badge above is providerWithdrawnAmount, the portion providers have actually pulled to their wallets so far via claimForProvider(); the rest is held in the contract pending claim. Tokens served is the provider-attested input + output counts from the closeoutReceipt (7-field format). Provider-signed, so self-reported and not independently verified, but it is the same receipt basis Morpheus's own dashboard reports from. All-time attested: 14.2B. Morpheus publishes a higher all-time figure (~14.93B as of June 2026) that adds a clamped estimate for the Dec-2025/Jan-2026 free open beta, whose 5-field receipts predate on-chain token counts; we show only the receipt-attested total we can verify. View contract on BaseScan. Source code at MorpheusAIs/Morpheus-Lumerin-Node. Refreshed daily. Last updated 2026-06-27.

Top models, last 30 days

Attested input + output tokens per model, read from provider-signed closeout receipts and summed over the last 30 days. Model names resolved on-chain from the registry.

deepseek-v4-flash
1.50B
deepseek-v4-pro:web
782M
glm-5.2
740M
glm-5
428M
glm-5.1
388M
kimi-k2.6:web
343M
kimi-k2.5
333M
clause-opus-4.8:web
303M
deepseek-v4-pro
288M
glm-5.1-non-thinking
143M

Reading the on-chain openers

Inference demand on Morpheus is real and growing, measured from provider-signed receipts, but it is intermediated almost entirely through the API Gateway, and part of it is off-chain: gateway users who pay in credits emit no on-chain receipt. The on-chain session opener is the address that opened the session, which on the gateway path is a shared API Gateway service wallet aggregating many end users behind one address (THROUGHPUT_DATA_SOURCES.md, section 4). Because of that aggregation, on-chain opener counts measure gateway service wallets, not end users, so they cannot show whether end-user demand is concentrated or broad. End-user diversity is not observable from chain data by construction; it would need gateway or app.mor.org billing data held by the protocol.

Source: SessionClosed receipts on the Lumerin Diamond, Base. Tokens are provider-attested. As of 27 Jun 2026.

Native vs gateway: who opens the sessions

The throughput above is one number per day. Split it by the wallet that opened each session and you can see how the demand reaches the network. The API gateway at app.mor.org opens sessions for its customers under a few service wallets, while self-custody clients like OpenClaw, Node Neo, or a local c-node open their own.

The self-custody share has climbed from a few percent in February toward roughly a third by mid-2026. That is the decentralisation signal worth watching: more inference reaching providers without passing through one hosted front door.

Who opens the sessions

Monthly attested tokens split by the wallet that opened each session. API Gateway is app.mor.org opening sessions for its customers under a few service wallets. Native is everything self-custody: OpenClaw, Node Neo, a local c-node.

Native share
36%
Jun · up from 4.3% in Feb
5.27B2.63B0
Feb Mar Apr May Jun
Native (self-custody) API Gateway
Each session records the wallet that opened it. The gateway opens one session per customer request under a handful of service wallets, so those wallets carry a large, persistent share of volume; every other opener is a self-custody client. We rank openers by attested token volume and tag the gateway service wallets, then bucket the rest as native. The split reproduces the Native vs API Gateway series on tech.mor.org within a few points per month. This is a different cut from the funding split above: that one asks who pays for a session, this one asks who opens it. Token weight is the provider-attested input + output from the closeout receipt, so the pre-launch beta months carry no tokens here. Refreshed daily. Last updated 2026-06-27.

Supply health

The charts above show whether usage is growing. This shows whether the small provider set is keeping up, and how close it is to its ceiling. Read the two together: demand rising while latency stays flat is health; demand rising while latency climbs is the turn. Today demand settles well inside headroom, so treat this as a baseline rather than a live alarm.

Supply vitals

Last 60 days · 7d line vs 30d band
Median TTFT 4.2s -18% vs base
Inside baseline band
Mean TPS 26.3 +3% vs base
Inside baseline band
Dispute rate 0.06% -66% vs base
Fast line below the band for 3+ days
Provider count 8 +0% vs base
Inside baseline band
Each panel plots the daily series (faint), a 30-day rolling-median baseline band (dashed, shaded) and a 7-day rolling-median fast line. The signal is the fast line leaving the band and staying out for 3+ days rather than a single-day spike. Median TTFT and mean TPS are provider-attested from the signed closeout receipt; dispute rate is the share of sessions closed with closeoutType = 1. Windows are config constants, set to be retuned once the real variance in these thin series is known.

Saturation curve

Per provider, each point is one day: tokens served that day against that day's median TTFT. Points are binned by load and fitted with a monotone trend. A flat left region is healthy headroom; where TTFT turns upward is that provider's inferred practical ceiling. The most recent 14 days are highlighted.

0s 6s 12s 17s 7613K2.3M410.8M Daily tokens served (log scale) →
0xb399…e362 0x8b59…6302 ceiling ~104.0M/day 0x4362…0a32 ceiling ~492K/day 0x0102…bb03
Serving capacity is off-chain, so the upward inflection is an inferred practical ceiling rather than a hard limit read from the chain. With one to three active providers these curves track one or two machines: the signal is present but thin until demand rises. The load axis is settled on-chain demand only; gateway usage billed off-chain in dollars emits no on-chain receipt, so true load is at least this and the gap is not yet quantified.

Concentration and resilience

Two bottlenecks sit on different axes. Resilience is a single point of failure, present now at one to three providers and independent of demand. Capacity is latent, arriving only as demand approaches aggregate serving capacity. This panel measures resilience; the saturation curve measures capacity.

Provider count (0–9) Top-provider token share (0–100%)
0
single-provider days (last 60)
0.49
latest HHI concentration (0–1)
Models served by only one provider (last 30 days) · 70 in total
  • Arcee-Trinity-Large-Thinking
  • Arcee-Trinity-Large-Thinking:web
  • claude-fable-5
  • claude-fable-5:web
  • claude-opus-4.8
  • clause-opus-4.8:web
  • deepseek-r1-70b:tee
  • DeepSeek-V3.2
  • DeepSeek-V3.2
  • deepseek-v4-flash:web
  • deepseek-v4-pro
  • deepseek-v4-pro:web
  • +58 more
HHI is the Herfindahl index of daily token share across providers: 1.0 is a single provider, lower is more even. Single-provider days count days where only one address closed sessions. A model served by only one provider over the window has no fallback if that operator drops it.

Headroom read. The busiest provider with a discernible ceiling (0x8b59…6302) has run as high as 172% of its estimated serving ceiling: the daily load where its latency starts to climb (~104.0M tokens/day), so readings above 100% mean it has pushed past that comfort point on its busiest days. The figure is an off-chain estimate inferred from the latency curve rather than a hard cap. The network has 8 independent providers closing sessions on the latest day.

The Morpheus team's tech.mor.org publishes the raw throughput counts from the same receipts. This section adds the supply-health reading those counts skip: the saturation curve, and resilience kept separate from capacity. Refreshed daily. Last updated 2026-06-27.

Two metrics, two trust models

MOR earned is the bottom-line economic activity on the network, calculated trustlessly from session start, end, and price. Tokens processed is provider-attested: the input + output counts signed into the closeout receipt. Treat one as audited revenue, the other as a self-reported throughput figure.

Where the MOR actually comes from

About 96% of all sessions are opened with isDirectPaymentFromUser = false: the user locks collateral that gets returned to them at session close, and the provider is paid out of the protocol's fundingAccount treasury on Base. The treasury is funded by the 24% compute pool slice of MOR emissions. Only 4% of sessions use the direct-payment path where the user's own MOR ends up with the provider. Most inference today is subsidised by emissions, not paid for by users. The runway implications are spelled out next.

The gateway's dollar side sits outside all of this. Card and crypto payments settle off-chain with whoever runs the gateway, so any decision to buy and stake MOR with that revenue is the operator's own, and the protocol mandates nothing of the sort. Treat gateway fiat as off-chain operator revenue that the protocol neither sees nor depends on, and look for structural MOR demand only where the chain can actually show it.

Sustainability headroom

As of writing, the on-chain compute pool surplus (getComputeBalance) is around 2.6 million MOR. The contract caps daily settlement at 1% of that surplus, around 26,000 MOR per day. Demand has been settling at 100-200 MOR earned per day. So today's usage is a tiny fraction of even the daily cap, and the cap itself is a tiny fraction of the accumulated reservoir. The runway is years, not months, at current demand.

A caveat about the early window

The Morpheus API ran an open beta from contract redeploy in December 2025 until 2 March 2026, when it went officially live. Across the beta, providers quoted near-zero prices by design, so MOR earned is essentially flat across the entire pre-launch period regardless of session volume. Some of the provider-signed token counts from the beta are also clearly test traffic. Headline numbers above sum the last 60 days, which falls almost exactly inside the post-launch era and keeps them representative of paid-era activity. Lifetime totals across the full dataset (including beta) live in the underlying JSON.

Provider concentration

The provider count is the part to watch on the recent end of the chart. Across the full deployment-to-now span, only a small number of distinct addresses have ever closed a session, and most days run with two or three. That's consistent with the bootstrap-phase reality the Morpheus project review describes; the marketplace works, but it's narrow. Pre-redeploy activity at the prior contract address is out of scope for this MVP.

The other side of the same coin: stake-for-access

The session settlement chart above is the use side of the network. The chart below is the access side: MOR locked into the Morpheus Marketplace API builder subnet to earn daily credits at app.mor.org. Same protocol, two different demand signals. Settlement counts work delivered. Stake counts intent to use the API.

Stake-for-access: Morpheus Marketplace API subnet

MOR locked into the official builder subnet that drives daily credit allowance at app.mor.org. Staked MOR stays in the contract; users earn credits proportional to their share of the subnet's total stake.

Total staked
163.5K MOR
9% of all builder subnets
Stakers
67
302 deposits / 69 withdrawals all-time

Total staked, end of day

On-chain · indexed from Base
163.5K81.7K0
25 Apr 7 May 20 May 1 Jun 15 Jun 27 Jun

Top builder subnets by stake

Mor.org-BASE 276.3K MOR (10)
MOR-StakeVault-Base 176.4K MOR (21)
Morpheus Marketplace API 163.5K MOR (67)
Lumerin Inference Node 160.2K MOR (2)
MySuperAgent 143.4K MOR (24)
MOR Deployers 111.1K MOR (2)
ShelterGuard 82.0K MOR (1)
Protection and Capital Incentive 80.8K MOR (3)
Indexed from UserDeposited and UserWithdrawn events on the BuildersV4 contract at 0x42bb…f4b9, filtered to the Morpheus Marketplace API subnet (0x415471125cc4…). All-subnet snapshot sourced from dashboard.mor.org. Refreshed daily. Last updated 2026-06-27.

Stake here doesn't pay providers; it just measures your share of the daily credit allocation. The MOR that ends up paying providers comes from the protocol treasury, as covered above. So this chart and the inference activity chart together describe the full economic picture: how much MOR is parked for access vs how much is being settled per session. Both are tracked daily.

Sources and reproducibility

If you want to use this dataset in your own research or writing, please attribute Own Your Mind and link back to this page.

For the full Morpheus project review (Freedom and Returns scoring, tokenomics, participation guide), see the Morpheus project page. For the MOR tokenomics deep-dive (Power Factor, demand-subsidy mechanics, burn under MRC43), see How MOR Actually Works. For every builder subnet ranked with trust signals, see the Morpheus subnet directory.