Kinds of data
Source code
Private repositories with their history: commits, reviews and the tickets they answer.
- Public price
- $934 to $24.2K
per repository
- Demand
- Split
- Who buys
- Coding-data vendors, environment builders, labs
The market rate for source code, by side
How the rates are computedThe median where three or more figures exist, otherwise the figures we hold. Each side is a different point in the market: what the people who make it are paid, what holders are offered, what catalogues ask, what was paid.
$130
per task, median of 7 figures; middle half $99 to $730
$118
per hour of work, median of 4 figures; middle half $92 to $151
$1,478
per repository, median of 6 figures; middle half $934 to $24.2K
$0.00004 to $0.0001
per token, from 2 figures
$1,000
per pull request, from 1 figure
$1,579
per repository, from 1 figure
$99 to $65.8K
per repository, from 2 figures
$0.0025
per record, from 1 figure
$8.94
per record, from 1 figure
$0.5
per record, from 1 figure
What source code has sold or been offered for
The whole price board- Review and upgrade of one real-world coding, data or automation scenario used to train AI assistants (Python, JavaScript, SQL, n8n, Make, Zapier), Vetto Coding, Data and Automations ExpertsPriceUS$ 30 per approved task
per approved task, about 40 minutes each
WhenOct 2026What to knowVetto (a Brazilian-founded expert network) pays the reviewer per approved task in local currency; its own effective rate is about $45 an hour. Read on AITraining.jobs, which has the same owner as AIData.Broker, quoting Vetto's posting (posted about seven months before the read, so dated by the read day).SourceAITraining.jobs, Vetto job posting: Coding, Data and Automations ExpertsThe payer's own page
- Original SWE-bench style coding task authored from the contributor's own private codebase (maintainer-style issue, hidden fail-to-pass and pass-to-pass tests, reference solution), AfterQuery University SWE ExpertPrice$75 per task that is fully approved
per fully approved task
WhenOct 2026What to knowAfterQuery pays the contributor only for tasks that pass automated validation (up to five attempts), on top of $300 per codebase connected; the same terms appear in AfterQuery's SWE Experts posting dated 14 Aug 2026, which also states $150 an hour. Undated posting, read on 2026-10-14.SourceHandshake job posting by AfterQuery, University SWE ExpertThe payer's own page
- Access to a substantial private codebase the contributor owns, used as the basis for SWE coding tasks (AfterQuery University SWE Expert)Price$300 per codebase you connect and use as a basis for tasks
per codebase connected
WhenOct 2026What to knowPaid to the owner of the private codebase per codebase used, separate from the $75 per approved task; the lowest per-repository figure a named buyer publishes, against $10K to $100K+ at Turing's Lazarus programme.SourceHandshake job posting by AfterQuery, University SWE ExpertThe payer's own page
- A well-documented, production-grade private codebase licensed on HUD's DataVendor marketplacePriceup to $20,000 each
per codebase
WhenOct 2026What to knowHUD's current ceiling, up from 'up to $10K' in June 2026. The marketplace says codebases have licensed for this much; it takes a 20% platform fee at withdrawal, so the holder nets up to $16,000. Read from the markdown copy the docs publish, because the rendered page typesets its dollar amounts as math and scrambles them for any reader but a browser; unchanged on 15 Oct 2026.SourceDataVendor (HUD) docs, What you can sell (the page's markdown)The payer's own page
- Custom coding dataset for LLM fine-tuning, focused single-language (1,000 to 5,000 pairs), made to order by LXTPrice$20K for focused single-language datasets (1,000-5,000 pairs)
per project of 1,000 to 5,000 pairs
WhenOct 2026What to knowA vendor's published entry price for made-to-order code data; the page names it as where projects start, not a rate card. Undated page, read on 2026-10-14.SourceLXT, Custom Coding Datasets for LLM Fine-Tuning and Code Generation (FAQ)A vendor's published price guide
- Custom coding dataset for LLM fine-tuning, large multi-language and multi-task with execution-verified code, made to order by LXTPrice$150K+
per project
WhenOct 2026What to knowOpen-ended top of LXT's stated range; no size is given, so no per-pair figure. Undated page, read on 2026-10-14.SourceLXT, Custom Coding Datasets for LLM Fine-Tuning and Code Generation (FAQ)A vendor's published price guide
- Reproducible coding RL environment (feature implementation or bug fix with a deterministic verifier and a golden reference solution), micro1 Software Engineer (GitHub Contributor)Price$50-100/hour, ~15 hours per run
per hour, about 15 hours per task
WhenOct 2026What to knowPaid hourly; the per-task figure is the hourly band times micro1's own stated time per task, so it moves with how long a task really takes. A twin posting (Software Engineer, Open Source Contributor) states the same pay and time.Sourcemicro1 job board, Software Engineer (GitHub Contributor)The payer's own page
- Git history (commits and their messages) licensed by the token through Pangea's code licensing partnerPrice$40 /M tokens of git history
per million tokens of git history
WhenOct 2026What to knowAn indicative standing rate, paid on top of the source-code rate for the same repository.SourcePangea.ai, Code Licensing referral partner guideThe payer's own page
- Private codebases licensed through Pangea's network to date: total paid to ownersPrice$600,000 across 380 codebases
in total, across 380 codebases
WhenOct 2026What to knowThe intermediary's own count of realised payouts. The average, about $1,579 a codebase, sits inside its 'typical' tier and far below the five- and six-figure ceilings buyers advertise.SourcePangea.ai, Code Licensing referral partner guideThe payer's own page
- A private repository of deep depth, licensed non-exclusively through Pangea's code licensing partnerPrice$1,500â$2,000 per repo (deep)
per repository, per licence
WhenOct 2026What to knowWhat the owner is paid per licence by Pangea's unnamed partner; Pangea calls its figures deliberately conservative estimates, not guaranteed payouts. Non-exclusive, so a repository can earn this more than once.SourcePangea.ai, Code Licensing referral partner guideThe payer's own page
- A private repository of light depth, licensed non-exclusively through Pangea's code licensing partnerPrice$700â$1,000 per repo (light)
per repository, per licence
WhenOct 2026What to knowWhat the owner is paid per licence by Pangea's unnamed partner; Pangea calls its figures deliberately conservative estimates, not guaranteed payouts. Non-exclusive, so a repository can earn this more than once.SourcePangea.ai, Code Licensing referral partner guideThe payer's own page
- A private repository of typical depth, licensed non-exclusively through Pangea's code licensing partnerPrice$1,000â$1,500 per repo (typical)
per repository, per licence
WhenOct 2026What to knowWhat the owner is paid per licence by Pangea's unnamed partner; Pangea calls its figures deliberately conservative estimates, not guaranteed payouts. Non-exclusive, so a repository can earn this more than once.SourcePangea.ai, Code Licensing referral partner guideThe payer's own page
- Private source code licensed by the token through Pangea's code licensing partnerPrice$100 /M tokens of source code
per million tokens of source code
WhenOct 2026What to knowAn indicative standing rate, the owner's side with nothing deducted; final value depends on the partner's analysis and buyer demand. A job-board copy of the same rates names Gramian Consulting Group.SourcePangea.ai, Code Licensing referral partner guideThe payer's own page
- AtCoder competitive-programming dataset (problems, test data, accepted and failed user solutions) licensed to AI companiesPricebudget is $100,000 or more
minimum per licence
WhenOct 2026What to knowSeller's own minimum order across four tiers (problems only, starter, all accepted, everything); full price list not yet published, and licensing of user code began August 2026.SourceAtCoder DatasetsThe payer's own page
- GitHub repositories dataset (3.4M+ records: repository attributes, languages, sizes, contributions, issues, pull requests, forks) sold by Bright DataPrice$250 for 100K records (approximately $0.0025 per record)
per 100,000 records (one-time)
WhenOct 2026What to knowPublic repository metadata, not licensed source code; a floor for what scraped public code facts cost.SourceBright Data, GitHub datasetsA catalogue or marketplace listing
- Data-structures-and-algorithms code repositories with programming solutions, price for an initial batch of 1,000 repositoriesPrice$98,700
per initial batch of 1,000 repositories
WhenOct 2026What to knowAsking price on the Opendatabay marketplace (USD since 1 Oct 2026, converted from GBP at $1 = GBP 0.76); seller-stated volumes and rights, not verified by us. Small exercise-style repositories, not production systems. Listed 4 Aug 2026.SourceOpendatabay, 12K+ DSA Codebases & 64K+ Programming Solutions Dataset (Infobay.AI)A catalogue or marketplace listing
- Legacy software codebases across 20+ industries (400+ codebases, 204M+ lines, 420K+ commits in the full set), price for an initial batch of 20 million linesPrice$196,100
per initial batch of 20 million lines of code
WhenOct 2026What to knowAsking price on the Opendatabay marketplace (USD since 1 Oct 2026, converted from GBP at $1 = GBP 0.76); seller-stated volumes and rights, not verified by us. The repository count in the batch is not stated, so no per-repository figure. Listed 4 Aug 2026.SourceOpendatabay, 400+ Legacy Codebases Dataset (Infobay.AI)A catalogue or marketplace listing
- 240+ private commercial repositories with full git history (300,000+ commits, ~42M lines, 44,000+ merged MRs with descriptions, ~463M tokens), de-identifiedPrice$263,200
per dataset of 240+ repositories
WhenOct 2026What to knowAsking price on the Opendatabay marketplace (USD since 1 Oct 2026, converted from GBP at $1 = GBP 0.76); seller-stated volumes and rights, not verified by us. At about $6 per merged MR it lists two orders of magnitude under Turing's $1,000 per PR offer. Listed 8 Sep 2026.SourceOpendatabay, Commercial Source Code & Engineering DataA catalogue or marketplace listing
- A private repository bought by Turing's Project Lazarus for AI training (all languages, priced on quality and complexity)PriceRepos typically pay $10kâ$100k+ depending on quality and complexity.
per repository
WhenOct 2026What to knowThe buyer's own seller-facing statement, broader than the legacy-only $10K to $100K+ figure on the board; the same page prices modern codebases at $1K per PR up to $10K and pays up to $1M for a whole company's record.SourceTuring, Project Lazarus (seller page)The payer's own page
- One self-contained database engineering task for AI agents (engine implementation, production diagnosis or application migration) with a reproducible repository, reference solution, tests and grading criteria, Mercor Database Systems Evaluation Expert pilotPrice$2.2K per task
per task, about 25 hours of work
WhenOct 2026What to knowA pilot for the first three qualifying applicants (applications close 16 Oct 2026, task due 30 Oct 2026), each completing one task; at the stated 25 hours it is about $88 an hour. Highest per-task coding rate on Mercor's board on the day read.SourceMercor job board, Database Systems Evaluation ExpertThe payer's own page
- Code Agent Trajectories: 551 verified PASS trajectories from an autonomous code agent in a sandboxed Python environment (reasoning, tool calls, final answer; JSONL, 5.63 MB), sold by Jurgen's Data Lab on GetlyPrice$100.00
per dataset of 551 trajectories
WhenSep 2026What to knowMachine-generated (an agent's runs judged PASS by an LLM), sold by a one-person storefront with commercial use allowed and no resale; the bottom of the market. The same seller lists a 177-trajectory PASS plus reasoning subset at $150 and a 924-entry bundle of its three sets at $250.SourceGetly, Fine-Tuning Datasets: Code Agent Trajectories, 551 Verified PASS (Jurgen's Data Lab)A catalogue or marketplace listing
- Premium: 177 verified PASS agent trajectories that also carry full chain-of-thought reasoning (JSONL, ATIF-v0.2, 1.3 MB), sold by Jurgen's Data Lab on GetlyPrice$150.00
per dataset of 177 trajectories
WhenSep 2026What to knowMachine-generated (an agent's runs judged PASS by an LLM), commercial use allowed, no resale of the file. The top-quality subset of the seller's 551-trajectory set (held at $100); costs 1.5 times the larger set for 177 entries, so the reasoning text is what is priced. The same seller also lists a 196-entry chain-of-thought set at $75 and a bundle of all three sets at $250; both repackage these same trajectories, so they are not counted as separate figures.SourceGetly, Jurgen's Data Lab: PremiumA catalogue or marketplace listing
- Software engineering expert pool (writing engineering problems with reference answers, grading model-written code)Price$70 to $150 an hour
per hour
WhenSep 2026What to knowMercor's own summary of what its software-engineering projects have paid; a standing pool, not one project.SourceMercor job board, Software engineering expert poolThe payer's own page
- A failed startup's codebasePriceTens of thousands of dollars
per codebase
WhenSep 2026What to knowOne buyer bought five to ten of them. Code alone, without the tickets and discussion around it, is the cheapest company record.SourceFortunePress report
- Application user recording real work in Visual Studio Code on macOS (computer-use data)Price$55-$65 per hour
per hour
WhenSep 2026What to knowOne of a family of per-application listings (PyCharm, Android Studio and others at the same rate); the worker supplies the machine and licence.Source - A private repository, graded by the platformPrice$5,000 baseline
per repository
WhenSep 2026What to knowTroveo's guide citing an unnamed codebase-buying platform's listed baseline, adjusted after grading. Production history, linked tickets and reviews, clean rights and an unusual domain raise it.SourceTroveo, How to sell source code and workflows to AI labsThe payer's own page
- Data-Analysis-Platform: one private multi-service enterprise Git repository with full history (5,683 commits, 1,474 tracked files, about 108,761 lines; Terraform, Python backend, IAM, SQL migrations, API tests), sold by Octobit8 on OpendatabayPrice$65,800
per repository (one repository, about 10.5 MB with history)
WhenSep 2026What to knowA seller's asking price on an open marketplace; the commit history spans about two years (Feb 2023 to Jan 2025), listed 10 Aug 2026, updated 2 Sep 2026. Opendatabay converted all GBP prices to USD on 1 Oct 2026 at $1 = 0.76 GBP, so this figure is a conversion of the seller's GBP price. Licence grants non-exclusive AI training rights, no resale.SourceOpendatabay, Software Engineering & Code Data: Data-Analysis-Platform (Octobit8 (OPC) Private Limited)A catalogue or marketplace listing
- Codebase-exploration RL task on a production Python repository (Django, Twisted, Scrapy, SymPy, mypy, pip and others), Mercor Codebase Q&APrice$130 per approved task
per approved task
WhenAug 2026What to knowSame project and rate as the Go, C and TypeScript tracks; listing active when read. Mercor posts the same $130 per approved task for its Go, C, TypeScript tracks (read 2026-10-14), so scarcer systems languages earn no premium; counted once because it is one programme at one rate. Closed: Mercor shows the posting no longer accepting applications (read 2026-10-17).SourceMercor job board, Software Engineer, Python: Codebase Q&AThe payer's own page
- Legacy codebase migration expert designing and grading migration tasks for a frontier AI labPrice$200 per hour
per hour
WhenAug 2026What to knowThe top hourly rate on the board for code work; legacy migration (COBOL, old Java and .NET) is also what Turing pays most for as codebases. - Frontier Code Agents project for an unnamed AI research lab: use AI coding agents to complete and evaluate realistic engineering tasks, review model-generated code for bugs and edge cases (Mercor, Backend Engineer, Coding Agent Experience)Price$400 per accepted task
per accepted task
WhenJun 2026What to knowOne programme at one rate: the Data Engineer, ML Engineer and DevOps/SRE/Cloud Engineer postings of the same project (all created 2026-06-15) state the same $400 and 2 to 3 hours, and the Software Engineers (Coding Agents Experience) posting of 2026-01-13 states $400 and 1 to 3 hours. All are closed (no longer accepting applications) when read; the figure sits in each posting's own description data, while the visible page now shows Mercor's standing $70 to $150 an hour band. About three times Mercor's $130 codebase Q&A task.SourceMercor job board, Backend Engineer (Coding Agent Experience)The payer's own page
- One prompt-and-code-review task (writing a prompt, A/B comparing two models' code and picking the better) for Claude Code fine-tuning, Snorkel AI's project Marlin for AnthropicPriceup to $280 per task
per task, typically about an hour
WhenJun 2026What to knowTwo contractors' account to Business Insider, relayed by a news summary because the Business Insider article was not fetched; a ceiling, not a published rate. At about an hour a task, up to about $280 an hour, far above the $30 to $130 per-task rates platforms post for comparable review. - Code repository seeded with an OSS-fuzz-style vulnerability plus a validated patch, AfterQuery Cybersecurity ExpertPriceEarn $300 per accepted task. Each successfully accepted task pays $50
per accepted task
WhenMay 2026What to knowThe posting states two figures for the same accepted task, $300 in its headline and $50 in its next sentence, plus bonuses up to $180 per repository; the range keeps both. The listing's structured pay also shows $75 an hour.Source - Legacy private codebase (COBOL, .NET, Java) licensed to Turing for AI trainingPrice$10K to $100K+ depending on complexity
per codebase
WhenMar 2026What to knowTuring's own published buy-side price guide for Project Lazarus; the top is open-ended and the final price depends on complexity.SourceTuring (AGI Advance newsletter, Project Lazarus)A vendor's published price guide
- Modern private codebase priced by pull request (Turing Project Lazarus)Price$1,000 per PR, up to $10K per codebase
per pull request (capped per codebase)
WhenMar 2026What to knowModern codebases are paid per PR with a $10K ceiling per codebase, far below the legacy-language range from the same buyer.SourceTuring (AGI Advance newsletter, Project Lazarus)A vendor's published price guide
- 258k test-verified coding-agent trajectories across 1,655 repositories (CoderForge-Preview, generated synthetically)Pricetotal cost of generating ... was $130k
per dataset (about $0.50 per trajectory, derived)
WhenFeb 2026What to knowA production cost at Together's own API prices, released free; it marks a floor for synthetic coding traces, not a sale price.SourceTogether AIA vendor's published price guide
- Coding-benchmark task work in React, TypeScript or JavaScript: curate issues, solutions and test suites from real-world repositories for AI labs (Mercor, Frontend Software Engineer), hourly base plus a bonus per approved taskPrice$90 per hour plus lucrative bonus per approved task (1 task takes approximately 1 hour to complete); Median average pay inclusive of bonuses is $200/hr
per hour (base), and median pay per hour including per-task bonuses
WhenNov 2025What to knowMercor states an hourly base plus an unstated bonus per approved task of about an hour; the $200 is its own median hourly pay including bonuses. Closed (no longer accepting applications) when read; one-month, 15 to 20 hours a week project.SourceMercor job board, Frontend Software Engineer (React, TypeScript or JavaScript)The payer's own page
- Professional developers' code snippets, edit history, prompts and terminal usage shared with JetBrains for AI trainingPricefree All Products Pack subscriptions for one year ... currently priced at $979.00 per user/year
per developer per year (paid in software licences)
WhenOct 2025What to knowPayment in kind, not cash, to selected organisations that opt in; the closest public price for real-world professional coding traces per seat.SourceThe RegisterPress report
What yours might fetch
A first answer or two gives a range. The full valuation asks the rest and shows where every figure comes from.
What is your data worth?
$3,700to$25,000
On two answers, against 3 public prices. An indication, not an offer.
- No data upload
- No account
- Free
Why labs pay for it
Public code is already in every model. What is not public, and above all the history that shows how a change was requested, argued over and shipped, is what teaches a model to work the way engineers do.
SplitOrdinary web and app code is now a commodity with at least seven intermediaries chasing it (2026). Scarce languages and code tied to its tickets still command a premium.
- Software companies
- Closed startups
- Agencies and independent developers
- Industrial and embedded teams
- A product's full repository history
- Code in scarce languages: COBOL, Fortran, Verilog, ABAP, PLC
- Internal tools nobody outside has seen
- Repositories with linked issues and reviews
- Reviews and tickets linked to the commits
- A scarce language or an unusual domain: embedded, scientific, industrial, finance
- Years of real production history
- Code a buyer can actually run
- Common web or app code with no history
- Generated code, duplicates and third-party libraries, which buyers strip out first
- Open-source licences or client ownership over parts of it
Code written for a client usually belongs to the client. Open-source components keep their own licences, and a contractor's work is yours only if the contract says so.
Single repositories are best sold direct to the platforms that grade and list them; we say so and point you there. We take on codebases that come with their tickets, their history or a scarce language, and whole-company sales.
Who buys it
Turing (Lazarus)
"A single payout of up to $1M" for an operating record. For code repositories its process page (lazarus.turing.com/process, read 16 Oct 2026) says "Repos typically pay $10kâ$100k+ depending on quality and complexity".
Ooak
"Earn up to $1M" (raised from $300K by 15 Oct 2026). Deals to date run from $20K to several hundred thousand dollars.
Labelbox (Prism)
"Smaller targeted datasets starting around $50K and large enterprise partnerships reaching $1M+".
Troveo
"AI labs and startups are actively paying six figures for company data exports", and Troveo pays the holder "on every sale, not just the first".
Corpus Bridge
It cites licences of about $10,000 to several hundred thousand dollars for ordinary companies.
SimpleClosure AssetHub
"Tens to hundreds of thousands of dollars for unique assets". Early deals were reported at $10,000 to $100,000.
Datacurve (Shipd)
Example bounties in the sample panels on its homepage run from $80 to $220 (read 16 Oct 2026; 2,847 registered contributors on its live counter). The panels show sample issues and challenges, not a rate card: "New quests launch every month across dozens of STEM categories".
Mercor
No figures. "Compensation is based on data volume, the tools you connect, and the depth of your data."
Where source code comes from
The whole supply map2,635,872
As of 16 October 2026, there are 2,635,872 apps on Google Play and 2,718,308 on the Apple App Store, each backed by a private codebase.
Reaches buyers: Direct deal (Google's confidential Play developer code pilot reported June 2026); otherwise not reaching buyers
42matters, Google Play vs iOS App Store | Store Stats for Mobile Apps, 16 Oct 2026
67.53TB
The Stack v2 contains over 3 billion files in 600+ languages, totalling 67.53TB uncompressed (about 900B training tokens per the card's table).
Reaches buyers: Open download (gated, terms of use)
BigCode on Hugging Face, bigcode/the-stack-v2 dataset card, Feb 2024
775-850 billion lines
A Micro Focus global survey estimated 775 to 850 billion lines of COBOL in daily use.
Reaches buyers: Not yet reaching buyers
IDM, COBOL Market 3X larger than Previously Estimated: Survey, 11 Feb 2022
81.5%
In 2025, 81.5% of contributions on GitHub (4.97B) happened in private repositories, though 63% of all repositories were public.
Reaches buyers: Not yet reaching buyers (private, owner-controlled)
630M
GitHub hosts 630 million repositories. 121 million were added in Octoverse year 2025 (Sep 2024 to Aug 2025), including 58 million new private repositories (up 33%).
Reaches buyers: Public: open download or scrape; private: not yet reaching buyers
nearly 500 million lines of code
The companies that offered assets on AssetHub since its April launch represent nearly 500 million lines of code across 3,000 repositories, with 1.7 million commits and 200,000 pull requests.
Reaches buyers: Wind-down service marketplace
What a buyer's counsel will ask
13 questions apply to source code, and 5 of them block a sale until settled. Most data that fails to sell fails here, not on price.
- Does the company selling the data own it, and has someone with authority approved the sale?
Blocks a sale until settled
- Does the data include personal information of employees, contractors or job applicants in California, and have they been told about the sale and given a way to opt out?
Blocks a sale until settled
- For staff in the EU or UK, what legal basis lets you reuse the personal data in their work (messages, documents, code history, recordings) for a sale, and is that new purpose compatible with why you collected it?
Blocks a sale until settled
- If the company is closing, insolvent or in bankruptcy, who now has legal authority to sell: the board, an assignee for the benefit of creditors, a bankruptcy trustee, or a Dutch curator?
Blocks a sale until settled
- Does the data include controlled technical information, such as semiconductor designs, aerospace or defence engineering, or advanced encryption source code?
Blocks a sale until settled
Hold source code? Describe it and we take it to the buyers.
In this topic
What sells
- Start hereWhich data do AI labs want, and who holds it?Each kind of data the market prices, the businesses that hold it, what labs are asking for now and what they turn away.
- GuideWhat Are RL Environments? Who Builds, Buys and Prices ThemAn RL environment is a sandbox where an AI agent attempts tasks with tools and a verifier scores each try. Who builds them, who buys them and what they cost.
- GuideHow to Sell Videos to AI Companies: Prices, Buyers and RightsAI companies license video by the hour or minute. What buyers pay, which footage sells, and the rights you need in every frame before you sell a library.
- PageWhere data comes fromWho holds each kind of data, how much exists and how it reaches buyers.
- RegisterCompanies closingCompanies closing, sold or gone dark, and the records they leave.