Datacircle

Best weather forecast API in 2026

Last updated:

Weatherbit, WeatherAPI.com and MET Norway are the best weather forecast API providers in our benchmark. We benchmarked 9 providers and rank all 9 by score. For the three we score highest, we show each one's top scoring plan, each part of the plan's score and what 1,000 places cost. We rank 165 plans by price. OpenWeather's £370 monthly plan is the cheapest paid plan per 1,000 places: $0.00049.

Every weather forecast API provider we score, ranked

  1. Weatherbit, $495 monthly, Business: Overall score 62.4, Data score 27, $0.0083 per 1,000 places
  2. WeatherAPI.com, $7 monthly: Overall score 61.7, Data score 38, $0.0023 per 1,000 places
  3. MET Norway, free: Overall score 61.0, Data score 18, $0.00 per 1,000 places
  4. Visual Crossing, pay as you go: Overall score 58.2, Data score 60, $0.10 per 1,000 places
  5. Open-Meteo, €160 monthly, API Professional (10 million calls): Overall score 56.9, Data score 37, $0.07 per 1,000 places
  6. Bright Sky, free: Overall score 51.4, Data score 9, $0.00 per 1,000 places
  7. API Ninjas, $299 monthly: Overall score 41.5, Data score 9, $0.03 per 1,000 places
  8. Strale, top-up from €10: Overall score 32.8, Data score 4, $56.27 per 1,000 places
  9. ReefAPI, $15 monthly: Overall score 29.0, Data score 19, $1.50 per 1,000 places

  1. #1 Weatherbit: Weather forecast API

    Weatherbit, $495 monthly, Business

    It returns all 6 key fields we check: Temperature, Precipitation, Wind, Conditions, Humidity, Sun.

    Overall score
    62.4
    $ per 1,000
    $0.0083
    Price score
    83
    • $ per 1,000 places the provider has no forecast for: $0
    • Plans of this API we rank by price: 165
    • Plans cheaper per 1,000 places than this plan: 32
    Data score
    27
    Cost to get started score
    62
    • How you pay: monthly
    • Least to pay before your first result: $495
    Bundle score
    83
    • Also pays for, from the same budget: Current Weather API · Hourly Forecast API (240 hour) · Minutely Forecast API (60 minute) and 2 more
    Performance score
    100
    • This call's median time on 30 October 2026, minus the baseline: 0.076 s, counted as 0.1 s
    • Strale, the fastest live provider we benchmarked: under 0.1 s, counted as 0.1 s
    • On 30 October 2026, we sent this provider all 20 places in our benchmark.
    • This plan scored 0.2741 out of 1 on those 20 places. That gives this plan a Data score of 27 out of 100.

  2. #2 WeatherAPI.com: Weather forecast API

    WeatherAPI.com, $7 monthly

    It returns all 6 key fields we check: Temperature, Precipitation, Wind, Conditions, Humidity, Sun.

    Overall score
    61.7
    $ per 1,000
    $0.0023
    Price score
    78
    • $ per 1,000 places the provider has no forecast for: the provider's pages don't say
    • Plans of this API we rank by price: 165
    • Plans cheaper per 1,000 places than this plan: 18
    Data score
    38
    Cost to get started score
    90
    • How you pay: monthly
    • Least to pay before your first result: $7
    Bundle score
    90
    • Also pays for, from the same budget: Realtime API · History API · Alerts API · Marine API · Search/Autocomplete API · IP Lookup API and 3 more
    Performance score
    35
    • This call's median time on 30 October 2026, minus the baseline: 0.282 s
    • Strale, the fastest live provider we benchmarked: under 0.1 s, counted as 0.1 s
    • On 30 October 2026, we sent this provider all 20 places in our benchmark.
    • This plan scored 0.3798 out of 1 on those 20 places. That gives this plan a Data score of 38 out of 100.

  3. #3 MET Norway: Weather forecast API

    MET Norway, free

    It returns 5 of the 6 key fields we check: Temperature, Precipitation, Wind, Conditions, Humidity.

    Overall score
    61.0
    $ per 1,000
    $0.00
    Price score
    100
    • $ per 1,000 places the provider has no forecast for: $0
    • Plans of this API we rank by price: 165
    • Plans cheaper per 1,000 places than this plan: 0
    Data score
    18
    Cost to get started score
    100
    • How you pay: free
    • Least to pay before your first result: $0
    Bundle score
    0
    • Also pays for: nothing else
    Performance score
    100
    • This call's median time on 30 October 2026, minus the baseline: 0.049 s, counted as 0.1 s
    • MET Norway, compact, the fastest provider we benchmarked that answers from a copy: under 0.1 s, counted as 0.1 s
    • On 30 October 2026, we sent this provider all 20 places in our benchmark.
    • This plan scored 0.1772 out of 1 on those 20 places. That gives this plan a Data score of 18 out of 100.

The Data score on weather forecast APIs

This score, out of 100, shows how much of the forecast a provider's call returns for a place. We send each place's latitude and longitude to the provider's call. For each place, we count how far ahead the forecast goes and how many records it holds. We also count the parts each record carries, such as the temperature and the wind. A record is one step of the forecast: an hour, a few hours, a day or a minute. We count only the records that end after our call. For each part of a record, we take the share of the call's own records that carry the part. A call with no day records scores 0 on each part of a day record. A call with no hour records scores 0 on each part of an hour record. A call with no minute records scores 0 on each part of a minute record. We count a part when a live call that offers the part returned it for the place. On a place where fewer than two live calls offer the part and return data, we count the part when any call returned it. For Days ahead, we divide by the furthest a live call went for the place. For each kind of record, we divide by the most records a live call gave for the place. We use all calls when only one live call or none found the place. We don't count a value given again in another unit. We don't count a value worked out from the date and the place, such as sunrise. We don't count the weather now, the air quality or the pollen that some calls also return. We don't count a heat index or a wind chill, which are other ways to work out Feels like. We don't check any value against another provider's, or against the weather the place got. Providers run different models, so their forecasts can differ for the same place. Then we average the scores of all places. Each place counts the same. We count a place as found when the answer holds a record that ends after our call and gives a temperature. When a provider didn't find a place, we count that place as 0. We compare a provider that ran on fewer places with the other providers on those same places. A provider can't score higher by running only on easy places. We lower the score when we test the call on fewer places than the benchmark has.

Best GTM APIs