yfinance

What it is

yfinance is a Python library that pulls price history, fundamentals, options chains and corporate actions from Yahoo Finance's internal endpoints and hands them back as pandas DataFrames. It costs nothing, needs no key, covers most of the world's exchanges, and is almost certainly what the tutorial you are reading uses.

It is also unofficial. There is no agreement behind it, no support, no uptime commitment and no rate limit you can plan against: neither yfinance's documentation nor Yahoo's API terms publish a number, and those terms reserve rate limits to Yahoo's "absolute and sole discretion". Every time Yahoo changes its endpoints, yfinance breaks for everyone simultaneously until a maintainer ships a fix. What the library does publish is what happens when you cross the invisible line — a 429 on a data request raises YFRateLimitError, "Too Many Requests. Rate limited. Try after a while.", while a 429 on the cookie-and-crumb handshake is downgraded to a warning and the request retried without it. Maintenance is real but volunteer — eight releases between mid-April and the end of August 2026, the last of them 1.7.0 on 26 August, and the merges behind them are signed by a community maintainer rather than by the author whose name is on the repo.

That combination makes it excellent for research and a poor foundation for anything with users.

Data & coverage

Global, delayed, and deep on daily bars — the docs state plainly that intraday data cannot extend past the last 60 days. Fundamentals are present and of uneven quality; treat anything surprising as a parsing artefact until a second source agrees. Beyond prices the library also wraps Yahoo's screener (predefined and custom queries), search, sector and industry data, news, and a websocket for streaming quotes — all of it as undocumented and unguaranteed as the rest.

Integrations

pip install yfinance is the whole install, and the dependency list is the interesting part: curl_cffi for browser-impersonating HTTP, pandas for the frames it hands back, peewee for the on-disk cookie and timezone cache, websockets and protobuf for the live stream. The one extra left is repair, which pulls scipy and scikit-learn for the price-repair code; the nospam extra that older tutorials still install was deleted in August 2026. The similar names on PyPI — yahoo-finance and yahoofinance — are unrelated packages by other authors.

Caching wrappers are now refused outright. Passing a requests_cache session raises YFDataException with "Caching sessions (e.g. requests_cache) are not supported", and any session that is not a curl_cffi or requests Session is rejected the same way, so the standard trick for softening the rate limit is gone.

A list passed to yf.download is not a batch request. In 1.7.0 it runs Ticker.history once per symbol, on up to twice as many threads as the machine has CPU cores, and a symbol that fails is logged as one of "N Failed downloads" and comes back as an empty column rather than an exception.

Downstream, OpenBB dropped it from its own repository in V5 (September 2026); openbb-yfinance 2.0.0, which requires yfinance>=1.3, is now published separately from its maintainer's own repository. vectorbt lists it in its full extra. In both cases the ceiling is Yahoo's, whatever the wrapper's own documentation promises.

Limitations

• No contract, no support, no SLA — by design, not by omission.
• Undocumented throttling. Nothing publishes a quota; you find the ceiling by hitting it and reading YFRateLimitError, and the caching sessions people used to blunt it are now refused.
• Breaks whenever Yahoo changes an endpoint.
• auto_adjust=False does not mean raw. Close is still split-adjusted, only not dividend-adjusted, so a script ported onto another vendor's unadjusted close reads every pre-split price several times too high.
• Spin-offs arrive as stock splits. 3M's April 2024 spin-off of Solventum is a 1.196 "split" in yfinance, with the dividend before it restated to match; see split and dividend history as data.
• Personal use only. The project's own README says so, and Yahoo's developer API terms forbid selling, sublicensing or deriving income from the APIs without Yahoo's written permission. The Apache-2.0 licence covers yfinance's code, not the data it returns. Training a model on it is the same question; Can you train a model on licensed market data? reads Yahoo's terms on it.
• It is a library, not an API, and it is catalogued here because that is where people look for it.

Alternatives

If the reason you are considering yfinance is money, the free tiers of Alpha Vantage are smaller but contractual. If it is coverage, the paid providers in the market data category start around $10 a month. The old shortcut of swapping the source in pandas-datareader is gone — 0.11.0 (23 June 2026) removed its Yahoo, Stooq and Tiingo readers — and replacing yfinance in a script covers the adapter that takes its place.