LBMA Precious Metal Benchmark Prices Data

It is the price discovery layer for physical bullion: the reference prices for gold, silver, platinum and palladium set in London auctions administered by the London Bullion Market Association, with the auctions themselves operated by ICE Benchmark Administration.

Each metal runs to its own clock. Gold is set twice daily, at 10:30 and 15:00 London time. Platinum sets twice daily at 09:45 and 14:00, and palladium follows in AM and PM rounds once the platinum auctions conclude. Silver takes a single noon fix. Every fixing lands in three currencies at once - US dollars, pounds sterling and euros, per troy ounce - with the sterling and euro lines indicative for settlement beside the dollar headline.

These are the numbers the rest of the market hangs off: contract settlement terms, ETF net asset values, mine revenue models and auditor marks all quote the same four benchmarks. Get a sample of this dataset and the fixings arrive as clean dated rows rather than chart widgets.

Rows exactly as they arrive:

# one observation = one metal x fixing round x day

metal_fixing : gold_am
date         : 2026-08-20
price_usd    : 4487.60
price_gbp    : 3288.96
price_eur    : 3833.76

metal_fixing : gold_pm
date         : 2026-08-20
price_usd    : 4482.95
price_gbp    : 3287.80
price_eur    : 3839.65

metal_fixing : silver              # single noon fix
date         : 2026-08-20
price_usd    : 66.745
price_gbp    : 48.92
price_eur    : 57.07

metal_fixing : platinum_am
date         : 2026-08-20
price_usd    : 1796.45
price_gbp    : 1317.70
price_eur    : 1535.61

window       : 2026-07-21 .. 2026-08-20   # 23 records per series, August 2026 vintage

Two things to notice. First, the AM/PM split survives into the data - the two gold prints above are the same day and nearly the same price, but they are distinct observations, and a model that blurs them loses the intraday anchor. Second, every row carries all three currencies together, so a dollar-sterling cross rate never has to be joined in from elsewhere.

Six core fields carry the spine, each defined in the table below. The design decision worth understanding before you build: one row is one metal-fixing-round, so the AM and PM sessions of the same metal are separate rows rather than columns - which keeps the panel rectangular when silver's single fix joins gold's two.

metal_fixing identifies the series, and date pins it to the London business day the auction ran. The three price fields hold the currencies side by side per troy ounce, and fixing_timestamp preserves the auction round label where it is carried, so a 10:30 gold print can always be told apart from a 15:00 one.

Beyond the core six sit supporting structures - trailing-week value arrays, period averages, pre-rendered chart images per metal, fixing, currency and period, and internal content flags. They ride along as additional fields on request: tell us which you need and they join the same feed rather than spawning a second pipeline.

• Geo: global benchmarks by construction - one world reference price per metal and session, quoted simultaneously in USD, GBP and EUR. There is no country cut to request because the fix is set once, in London, for everyone.
• Temporal: a rolling recent window of daily fixes ships as standard - 23 records per metal-fixing series observed in the August 2026 vintage, spanning 2026-07-21 to 2026-08-20 - with deeper historical spans scoped to your specification on request.
• Granularity: twice-daily AM and PM fixes for gold, platinum and palladium; a single noon fix for silver. The fix calendar follows London business days, so weekends and UK bank holidays carry no print by design rather than by gap.

One structural note: because the currency triple travels on every row, the same feed serves a dollar-based risk system and a sterling-reporting fund without any conversion layer in between.

API, files, or your warehouse. Daily, weekly, or hourly.

Pick the channel your stack already speaks and set the cadence to match the decision being fed. A treasury desk marking books at the London close wants both gold fixes and the silver noon print landed before reconciliation runs; a miner-revenue model may only need a daily pull; a monitoring layer can watch for the AM print landing and fire on movement. Cadence changes are a settings conversation, not a re-integration project.

Additional fields - the trailing-week arrays, period averages and chart renders - ride the same channel once scoped, so extending the feed never forks the pipeline.

Ranked by how directly one row settles the day job:

1. Treasury and trading desks. Contract settlement and position marking need a price both counterparties accept without argument - that is precisely what an auction benchmark exists to be, and the AM/PM pair covers a full London trading day.
2. Equity and credit analysts covering miners. Revenue lines keyed to realized gold, platinum and palladium prices stop being estimates when the reference print is the same number the offtake contracts cite.
3. Quant funds and backtesters. Dated AM and PM rows give an intraday anchor pair per session, enough to sanity-check higher-frequency strategies against the market's own rhythm without a tick-data budget.
4. Fund administrators, auditors and valuers. An end-of-day mark in three currencies from a named auction process is the defensible input a NAV or an audit trail wants.

Investors and quants get the market's own reference points as typed rows - AM and PM pairs per session, three currencies per row - to pin portfolio marks and benchmark their own price views against. Data scientists get a rectangular panel where one row is one metal-session-day, so feature engineering starts at analysis instead of untangling chart widgets. Developers and builders get the fixings on a cadence they set, arriving through whichever channel the stack already speaks.

Three things, stated up front. First, this is the benchmark view: one world price per metal and session, not quotes from individual dealers or venues - if you need dealer-level dispersion, say so and it gets scoped alongside. Second, the standard window is rolling and recent, so if your backtest reaches back years, name the span in the sample request and the deeper history gets attached to the same feed. Third, the currency trio is per-row by design; treat the sterling and euro lines as settlement-indicative companions to the dollar headline, not independent benchmarks.

Get a sample of this dataset, or browse the diversified metals mining data hub for the rest of the industry.