# Dune vs Nansen: SQL you write, or a label set you read One meters compute over 100+ chains for queries you write; the other sells a label set you read and cannot inspect. The units do not compare. *https://cryptomarkets.tools/compare/dune-vs-nansen* - https://cryptomarkets.tools/tools/dune.md - https://cryptomarkets.tools/tools/nansen.md ## Two products under one search term Both are filed under on-chain analytics, both sit on the same public chain data, and both put roughly seventy dollars a month between you and the useful part. That is where the resemblance ends. The category page splits this market into three kinds of product, and these two are two different kinds. **[Dune](https://cryptomarkets.tools/tools/dune) sells a query engine.** It indexes and decodes more than 100 chains into raw, ABI-decoded and curated tables, and hands you a SQL editor and an API over them. Nothing is answered for you: no supplied definition of an active address, no name on any wallet, no dashboard you did not build or borrow. **[Nansen](https://cryptomarkets.tools/tools/nansen) sells a labelled view.** It resolves addresses to entities — an exchange's hot wallet, a fund, one of its own Smart Money cohorts — and builds every screen and endpoint on that attribution. There is no SQL surface and no notebook. You read Nansen's screens and call Nansen's endpoints, and the questions you can ask are the ones Nansen has built a view for. So the first decision is not a price. It is whether the question you have is your own or one somebody else has already turned into a screen. ## What each one asks you to bring **Dune asks for SQL and time.** The ceiling is your own SQL and your patience with a schema the card describes as having a thousand tables in it, organised differently per chain. The editor's AI tools will generate, edit and debug SQL, but the schema is still yours to learn, and the time spent on it comes before the first answer you trust. **Nansen asks for a subscription and trust.** The work of deciding who owns a wallet has been done for you, and the price of skipping it is that you accept the result as given. Pro at 69 dollars a month (49 on the annual tab) is the plan that carries the labels; the free plan cannot do the thing the product is for, because its API side buys one common-labels lookup in total. The free side runs the opposite way from what most readers expect. Nansen, the hosted dashboard, has a free account that stays an account. Dune, the tinkerer's tool, runs nothing for free after a 14-day trial with 2,500 credits in total; from then on you can browse dashboards and queries and run none of them until you pay 75 dollars for Analyst. ## Chains, depth, and how late **Dune covers more chains, unevenly.** More than 100, by its own count, but the depth is the point. EVM chains get the full stack down to decoded contract calls and curated tables such as `dex.trades`; Solana has its own first-class shape; Bitcoin is blocks, inputs, outputs and transactions with no decoded layer; most of the rest are raw tables plus token transfers. Derivatives coverage is essentially Hyperliquid. **Nansen covers fewer, and labels fewer still.** It documents 38 chains and says support varies by endpoint: the labels endpoint accepts 20 of them. History is per chain and starts where Nansen onboarded it — Ethereum from July 2015, Solana from March 2020, Base from June 2023, newer chains from the day they arrived. A chain in Nansen's list is not necessarily a chain with names on it. **Freshness is published on one side only.** Nansen states seconds to a couple of minutes behind the chain, with some current-day responses served from a cache of up to five minutes. Dune publishes no figure and no SLA outside Enterprise; its card records the data as delayed and says to measure the lag yourself on the chain you care about. Neither is a feed to act inside a block on, and the category page says so of the whole group. ## Where a label comes from, and whether you can read the rule This is the axis the search term hides. On **Nansen**, a label is an object with a name, a category and a kind. Nansen puts the count at 500 million labelled addresses. The object carries no date, the endpoint returns the current set only, and there is no history — so two answers a year apart cannot be told apart, and Nansen lists label-history updates among the reasons a past result can change when you re-run it. The Smart Money cohorts are thresholds Nansen chose — minimum win rates, required ROI, realised-profit floors — over sets of roughly 850 all-time smart traders on EVM chains. Whether a given wallet belongs in one cannot be checked from outside. You are buying a judgement, and you cannot audit it. On **Dune**, there is no first-party label at all. The community tables in the catalogue are other users' work, not a maintained product with a revision date. A label on Dune is whatever your query says it is — which means the rule is in your SQL, where you, a colleague or a reviewer can read it, and it is wrong in ways you can find. That is the trade in one line: Nansen's labels are many and opaque, Dune's are absent unless you write them, and then they are yours. ## What the meter prices on each side Both bill API usage in something called a credit. The two credits are not the same unit. **A Dune credit is compute.** Dune bills in proportion to CPU, bytes scanned and engine time, and states that the cost appears after execution; there is no estimate in advance. The plan price is the allowance, bought at the plan's own overage rate: Analyst's 75 dollars for 4,000 credits is exactly the 1.875 dollars per 100 it charges past the allowance, Plus's 399 for 25,000 is exactly its 1.596, and the annual figures (65 and 349 a month) match the annual overage rates of 1.625 and 1.396 to the tenth of a cent. Moving up a plan buys cheaper credits and better export terms, not a bonus. Credits do not roll over, and the only defences against a runaway query are a per-execution cost cap and a monthly limit you set yourself. **A Nansen credit is an endpoint call at a fixed price.** Most Token God Mode and Profiler calls cost 1 credit, Smart Money endpoints 5, an address-labels lookup 100 for common labels and 500 for premium ones, an agent question 200 or 750. You know the price before you call. Pro tops the balance back up to 2,000 credits a month only if it has fallen below that — a floor, not a grant — which is twenty common-labels lookups or four premium ones if you spend on nothing else. The pay-per-call rail puts a dollar figure on a Nansen credit: one cent for a basic endpoint the subscription charges one credit for, five cents for a five-credit one. At that rate the Pro floor is about twenty dollars of one-credit calls, so most of the 69 dollars buys the app and the labels — and the labels are exactly what no rail sells by the call. There is no honest exchange rate between a Dune credit at roughly 1.4 to 1.9 cents and a Nansen credit at about one. One buys an unknown slice of a warehouse scan; the other buys a known response. ## Getting data out, and what you may do with it **Dune has more exits.** A REST API on every paid plan and the trial, metered in the same credits as the editor; first-party SDKs for Python, TypeScript and Go; a Trino connector for BI tools and a dbt connector with write access; CSV and JSON exports; scheduled queries that fire email or webhook notifications; and a first-party MCP server. The cost of an export depends on which Dune page you read — one credit per 1,000 or 5,000 data points in the pricing FAQ, 10 or 2 credits per MB on the API billing page — so price a large one by running a small one first. Results are capped at 32GB. An account-free route answers with an HTTP 402 and takes crypto per request, at a price Dune does not publish. **Nansen has the API and little else.** A REST API keyed by header, a first-party MCP server and a CLI, but no SDK; its card records the API as the only export path. Pay-per-call over x402 or MPP needs no key, excluding the labels endpoints. Alerts go to Telegram, Discord, Slack or an HTTPS endpoint of your own, optionally signed. **The licences cut the other way.** Every Dune self-serve plan, Plus included, reads *internal use only* on the vendor's own plan table; a commercial licence is Enterprise and unpriced. Nansen prohibits redistribution of address labels, Smart Money holdings and trades and both PnL leaderboards outright, wants its approval for a middle group, and allows most other endpoints with visible attribution. So the product with more exits lets less of what leaves be shown outside your organisation at any published price, and the product with fewer exits locks the labels but not everything around them. ## The cards were read on different days Nansen's pricing and coverage were both read on 19 September 2026. Dune's coverage was read the same day and its pricing a week later, on 26 September, after its 10 September move to a view-only free tier. Nansen's in-app plan grid sits behind a bot checkpoint and has not been read, so what a free Nansen account sees in the browser beyond one alert is not on its card. Dune's own pages disagree with each other on export pricing and on a free API rate limit; its card records both versions. ## Which one, and when **Take Dune** when you are building something — a dashboard, a recurring research series, a number you will have to defend — and you have the SQL and the hours. It is the only one of the two where the definition of every figure, a wallet's name included, is yours to read and change. Start on the 14-day trial, measure what your real queries cost in credits, and only then choose between Analyst and Plus; if the output has to reach anyone outside your organisation, that is an Enterprise conversation before it is a plan. **Take Nansen** when the question starts with "who" — which named entities hold a token, what the wallets Nansen rates are doing, alerts on Smart Money flows — and you will read the answer rather than rebuild it. Pro is what you are buying, and the label lookups inside it are scarce enough to budget. Watching labelled wallets tells you what those wallets did, on Nansen's say-so about who they are; it is not a method for anything beyond that. **The condition that decides it:** Nansen until the first time you need to show how a wallet got its name; Dune from that moment. A reader who needs both — names *and* a query surface — is describing neither of these, and the nearest card is [Allium](https://cryptomarkets.tools/tools/allium), a SQL product with a first-party label set and no published price. The labelled side has its own head to head in [Nansen vs Arkham](https://cryptomarkets.tools/compare/nansen-vs-arkham). ## FAQ ### Which one is cheaper? The headline prices are close and the units are not. Dune's cheapest plan that runs a query is Analyst at 75 dollars a month for 4,000 credits, and a credit there is compute whose cost per query you learn only after it has run. Nansen Pro is 69 dollars a month, or 49 on the annual tab, and each API endpoint has a fixed credit price you know before calling it. Decide which product shape you need first; the price follows from that. ### Can I use either without paying? Nansen, yes, within narrow limits. Its free plan is a real account with one active smart alert, 100 opening API credits and a daily top-up to 10, which buys one common-labels lookup in total. Dune, only for 14 days. Since 10 September 2026 a new account gets a trial with 2,500 credits in total, and after that free access is view-only, with no query execution, no API and no way to buy extra credits. ### Does Dune label wallets the way Nansen does? No. Dune has no first-party attribution layer. Community-maintained tables exist in its catalogue, but they are other users' work with no revision date. What Dune gives you instead is the ability to write the rule yourself in SQL, which is more work and the only version in which you can show how a wallet got its name. ### Which one is closer to the chain tip? Nansen, by its own description. It states live data runs seconds to a couple of minutes behind the chain, with some current-day responses cached for up to five minutes. Dune publishes no freshness figure at all and its card records the data as delayed. Neither is fresh enough for a decision that has to be made inside a block.