The Kimi K3.1 Pro launch produced the now-familiar on-chain reaction within hours: an X thread predicting decentralized providers would list the model at 30-80% off, a reply claiming Surplus already had it live in about four hours, and talk of confidential inference ahead of the November 4 open-weights drop. The verifiable part underneath: Surplus's own market page shows a live Kimi K3.1 Pro order book with 22 offers.
Key facts
- 22 offers, $0.11/M input best price — Surplus Intelligence's market page for Kimi K3.1 Pro lists 22 active offers, with the cheapest at $0.11 per million input tokens against a $2.00/$10.00 direct reference price — a listed discount of 99.08%.
- "~4 hours" — the claimed time from launch to Surplus listing, per @mac_eth's reply; the timing itself we could not verify.
- 30-80%+ off — @0xJeff's prediction for on-chain pricing, framed as expectation ("I can't imagine a world where..."), not a report of observed prices.
- ~83% mean discount — Surplus's self-reported mean realized discount on the model over the past day, across 14,374 eligible requests.
What's driving the conversation
@0xJeff's October 24 post argues the pattern will repeat: "I can't imagine a world where these models don't get immediately integrated by onchain providers and sold for 30-80%+ off." The post bundles three threads — Kimi K3.1 Pro matching frontier capability, local conversational models, and "renewed wave of interests for Confidential Inference" — and closes with "November is gonna be an exciting month for Onchain Inference." @mac_eth's reply supplies the same-day datapoint: "Kimi K3.1 Pro has been live on surplus for ~4 hours already." A follow-on from @beth2049_ argues TEE-based confidential inference is "the cleanest answer so far" for anyone touching trading or wallet data once open models get close to frontier quality.
An earlier @0xJeff thread from October 22 puts market context around the enthusiasm, naming Surplus alongside Phala, engyAI, Dolphin, and Venice as the "most undervalued" on-chain inference teams on a market-cap-to-token-usage comparison, citing Surplus token usage of "250B+ tok/day" — a self-report we have not verified. Engagement on the October 24 post is modest (28 likes, 0 retweets at time of extraction); the pattern, not the virality, is the story.
The substance
Two things check out against Surplus's own pages. The model is listed: surplusintelligence.ai/markets/kimi-k3.1-pro carries a live order book with 22 offers, the cheapest at $0.11 per million input tokens and $0.92 per million output against the $2.00/$10.00 reference price from Moonshot's launch — a listed best discount of 99.08%. Surplus's analytics frame also reports 28-day token volume of 5.5T input tokens and 110.7M routed requests, both self-reported platform metrics.
One nuance matters for anyone reading the discount numbers: Surplus's page reports a mean realized discount of about 83% and a median discount band of 95% over the past day — the realized average sits well below the headline 99% best offer, which is typical for order books where the cheapest offers are capacity-capped. None of these figures are independently audited; they are Surplus's own measurements of its own market.
What remains claim-only: the ~4-hour listing time (a single reply), the 250B+ tokens/day usage figure, and the general 30-80% framing, which was a prediction. Whether these discounts persist once the model's novelty fades, and what sellers' effective quality looks like at the cheapest offers, is not yet answerable from public data.
Why builders are watching
Time-to-listing is the interesting mechanic. Moonshot AI's Kimi K3.1 Pro went live on the Kimi API at $2/$10 per million tokens with open weights promised by November 4 — we covered the launch — and within roughly a day, an open marketplace had a 22-seller order book around it. That listing speed is a structural property of marketplaces built on resellers with existing serving capacity, and it is the argument on-chain advocates keep making against closed endpoints that integrate on vendor timelines. The decentralized inference explainer covers the mechanism; the router comparison openrouter-vs-morpheus shows the same dynamic on centralized aggregators.
The TEE thread is worth watching separately. If Moonshot ships full open weights on November 4 as promised, anyone can serve the model — and the question shifts from capability to privacy guarantees, where trusted-execution-environment serving is the current front-runner for prompts that touch wallet or trading data. Watch whether Surplus, Phala, and peers ship confidential-inference options for the weights drop; that would turn this week's discourse into product.
Questions
- Did Surplus really list Kimi K3.1 Pro within 4 hours of launch?
- The ~4-hour claim is a reply from @mac_eth on X and we could not independently verify the timing. What is verifiable: Surplus Intelligence's market page for kimi-k3.1-pro shows a live order book with 22 offers, a best listed input price of $0.11 per million tokens against a $2.00 direct reference price, and a best listed discount of 99.08%.
- What is disputed about the Surplus Kimi K3.1 Pro listing?
- The 30-80% discount figure in the original X post was a prediction, not a measurement, and the ~4-hour listing time is unverified. Surplus's own page reports a median realized discount of about 83% and a median discount band of 95% for the model over the past day — different numbers from the thread's range, and both are platform self-reports.
- What is TEE-based confidential inference?
- Trusted-execution-environment inference runs a model inside a hardware-isolated enclave so the operator cannot read the prompts or outputs passing through it. X commenters this week argued TEE inference becomes more attractive once open models approach frontier quality, because anyone touching trading or wallet data may want frontier-class capability without exposing prompts to the operator.
Sources
About DeAI
DeAI is an independent publication covering open-weight AI models, private inference, and decentralized infrastructure — the tools for running AI you actually control. We test providers on price, privacy, and refusal behavior and publish the numbers, not the vibes. DeAI is powered by Morpheus (mor.org), a decentralized inference marketplace, and covers it on the same terms as every other provider.
Powered by Morpheus and StrandCMS
Morpheus is a decentralized inference marketplace, covered on the same terms as every other provider — we rank it wherever the data lands. StrandCMS is the open-source, agent-first framework this site is built on.