diff --git a/CHANGELOG.md b/CHANGELOG.md
index d33e2bc..4d02509 100644
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -17,6 +17,7 @@ sort: 15
 * FEATURE: preserve the order of time series passed to [limit_offset](https://docs.victoriametrics.com/MetricsQL.html#limit_offset) function. This allows implementing series paging via `limit_offset(limit, offset, sort_by_label(...))`. See [this](https://github.com/VictoriaMetrics/VictoriaMetrics/issues/1920) and [this](https://github.com/VictoriaMetrics/VictoriaMetrics/issues/951) issues.
 * FEATURE: automaticall convert `(value1|...|valueN)` into `{value1,...,valueN}` inside `__graphite__` pseudo-label. This allows using [Grafana multi-value template variables](https://grafana.com/docs/grafana/latest/variables/formatting-multi-value-variables/) inside `__graphite__` pseudo-label. For example, `{__graphite__=~"foo.($bar)"}` is expanded to `{__graphite__=~"foo.{x,y}"}` if both `x` and `y` are selected for `$bar` template variable. See [these docs](https://docs.victoriametrics.com/#selecting-graphite-metrics) for details.
 * FEATURE: add [timestamp_with_name](https://docs.victoriametrics.com/MetricsQL.html#timestamp_with_name) function. It works the same as [timestamp](https://docs.victoriametrics.com/MetricsQL.html#timestamp), but leaves the original time series names, so it can be used in queries, which match multiple time series names: `timestamp_with_name({foo="bar"}[1h])`. See [this comment](https://github.com/VictoriaMetrics/VictoriaMetrics/issues/949#issuecomment-995222388) for more context.
+* FEATURE: add [changes_prometheus](https://docs.victoriametrics.com/MetricsQL.html#changes_prometheus), [increase_prometheus](https://docs.victoriametrics.com/MetricsQL.html#increase_prometheus) and [delta_prometheus](https://docs.victoriametrics.com/MetricsQL.html#delta_prometheus) functions, which don't take into account the previous sample before the given lookbehind window specified in square brackets. These functions may be used when the Prometheus behaviour for `changes()`, `increase()` and `delta()` functions is needed to be preserved. VictoriaMetrics uses slightly different behaviour for `changes()`, `increase()` and `delta()` functions by default - see [this article](https://medium.com/@romanhavronenko/victoriametrics-promql-compliance-d4318203f51e) for details. See [this issue](https://github.com/VictoriaMetrics/VictoriaMetrics/issues/1962).
 
 * BUGFIX: fix `unaligned 64-bit atomic operation` panic on 32-bit architectures, which has been introduced in v1.70.0. See [this issue](https://github.com/VictoriaMetrics/VictoriaMetrics/issues/1944).
 * BUGFIX: [vmalert](https://docs.victoriametrics.com/vmalert.html): restore the ability to use `$labels.alertname` in labels templating. See [this issue](https://github.com/VictoriaMetrics/VictoriaMetrics/issues/1921).
diff --git a/MetricsQL.md b/MetricsQL.md
index dc3912c..ea4bf73 100644
--- a/MetricsQL.md
+++ b/MetricsQL.md
@@ -92,7 +92,11 @@ See also [implicit query conversions](#implicit-query-conversions).
 
 #### changes
 
-`changes(series_selector[d])` calculates the number of times the raw samples changed on the given lookbehind window `d` per each time series returned from the given [series_selector](https://prometheus.io/docs/prometheus/latest/querying/basics/#time-series-selectors). Metric names are stripped from the resulting rollups. This function is supported by PromQL.
+`changes(series_selector[d])` calculates the number of times the raw samples changed on the given lookbehind window `d` per each time series returned from the given [series_selector](https://prometheus.io/docs/prometheus/latest/querying/basics/#time-series-selectors). Unlike `changes()` in Prometheus it takes into account the change from the last sample before the given lookbehind window `d`. See [this article](https://medium.com/@romanhavronenko/victoriametrics-promql-compliance-d4318203f51e) for details. Metric names are stripped from the resulting rollups. This function is supported by PromQL. See also [changes_prometheus](#changes_prometheus).
+
+#### changes_prometheus
+
+`changes_prometheus(series_selector[d])` calculates the number of times the raw samples changed on the given lookbehind window `d` per each time series returned from the given [series_selector](https://prometheus.io/docs/prometheus/latest/querying/basics/#time-series-selectors). It doesn't take into account the change from the last sample before the given lookbehind window `d` in the same way as Prometheus does. See [this article](https://medium.com/@romanhavronenko/victoriametrics-promql-compliance-d4318203f51e) for details. Metric names are stripped from the resulting rollups. This function is supported by PromQL. See also [changes](#changes).
 
 #### count_eq_over_time
 
@@ -124,7 +128,11 @@ See also [implicit query conversions](#implicit-query-conversions).
 
 #### delta
 
-`delta(series_selector[d])` calculates the difference between the first and the last point over the given lookbehind window `d` per each time series returned from the given [series_selector](https://prometheus.io/docs/prometheus/latest/querying/basics/#time-series-selectors). Metric names are stripped from the resulting rollups. This function is supported by PromQL. See also [increase](#increase).
+`delta(series_selector[d])` calculates the difference between the last sample before the given lookbehind window `d` and the last sample at the given lookbehind window `d` per each time series returned from the given [series_selector](https://prometheus.io/docs/prometheus/latest/querying/basics/#time-series-selectors). The behaviour of `delta()` function in MetricsQL is slighly different to the behaviour of `delta()` function in Prometheus. See [this article](https://medium.com/@romanhavronenko/victoriametrics-promql-compliance-d4318203f51e) for details. Metric names are stripped from the resulting rollups. This function is supported by PromQL. See also [increase](#increase) and [delta_prometheus](#delta_prometheus).
+
+#### delta_prometheus
+
+`delta_prometheus(series_selector[d])` calculates the difference between the first and the last samples at the given lookbehind window `d` per each time series returned from the given [series_selector](https://prometheus.io/docs/prometheus/latest/querying/basics/#time-series-selectors). The behaviour of `delta_prometheus()` is close to the behaviour of `delta()` function in Prometheus. See [this article](https://medium.com/@romanhavronenko/victoriametrics-promql-compliance-d4318203f51e) for details. Metric names are stripped from the resulting rollups. See also [delta](#delta).
 
 #### deriv
 
@@ -180,7 +188,11 @@ See also [implicit query conversions](#implicit-query-conversions).
 
 #### increase
 
-`increase(series_selector[d])` calculates the increase over the given lookbehind window `d` per each time series returned from the given [series_selector](https://prometheus.io/docs/prometheus/latest/querying/basics/#time-series-selectors). It is expected that the `series_selector` returns time series of [counter type](https://prometheus.io/docs/concepts/metric_types/#counter). Metric names are stripped from the resulting rollups. This function is supported by PromQL. See also [increase_pure](#increase_pure) and [delta](#delta).
+`increase(series_selector[d])` calculates the increase over the given lookbehind window `d` per each time series returned from the given [series_selector](https://prometheus.io/docs/prometheus/latest/querying/basics/#time-series-selectors). It is expected that the `series_selector` returns time series of [counter type](https://prometheus.io/docs/concepts/metric_types/#counter). Unlike Prometheus it takes into account the last sample before the given lookbehind window `d` when calculating the result. See [this article](https://medium.com/@romanhavronenko/victoriametrics-promql-compliance-d4318203f51e) for details. Metric names are stripped from the resulting rollups. This function is supported by PromQL. See also [increase_pure](#increase_pure), [increase_prometheus](#increase_prometheus) and [delta](#delta).
+
+#### increase_prometheus
+
+`increase_prometheus(series_selector[d])` calculates the increase over the given lookbehind window `d` per each time series returned from the given [series_selector](https://prometheus.io/docs/prometheus/latest/querying/basics/#time-series-selectors). It is expected that the `series_selector` returns time series of [counter type](https://prometheus.io/docs/concepts/metric_types/#counter). It doesn't take into account the last sample before the given lookbehind window `d` when calculating the result in the same way as Prometheus does. See [this article](https://medium.com/@romanhavronenko/victoriametrics-promql-compliance-d4318203f51e) for details. Metric names are stripped from the resulting rollups. This function is supported by PromQL. See also [increase_pure](#increase_pure) and [increase](#increase).
 
 #### increase_pure