1. What this system is

Control-IQ is a hybrid closed-loop AID algorithm developed by Tandem Diabetes Care. It runs on the t:slim X2 insulin pump and uses real-time glucose data from a Dexcom G6 or G7 CGM to make automated insulin delivery decisions. The key architectural feature is that it acts on two channels simultaneously: it adjusts the ongoing basal rate and, under certain conditions, delivers an automatic correction bolus.

The predictive element is central to how Control-IQ operates. Rather than responding purely to current glucose, the algorithm projects where glucose is likely to be in 30 minutes. If that predicted value meets the threshold criteria, it increases basal delivery to prevent the predicted rise, or reduces it to prevent a predicted fall. If glucose is predicted to remain above 10.0 mmol/L (180 mg/dL) after 30 minutes, an automatic correction bolus can be delivered — up to once per hour.

What makes Control-IQ distinctive

• Dual-channel automation. Both basal and bolus insulin are adjusted automatically — not just the basal rate. This gives the algorithm more tools to respond to glucose excursions than systems limited to basal modulation alone.
• 30-minute predictive horizon. Every insulin delivery decision is made on the basis of where glucose is predicted to be in 30 minutes, not just where it is now. This predictive window is a fundamental part of Control-IQ’s design.
• Sleep mode as a key lever. Sleep mode in Control-IQ operates differently from standard mode — it uses a lower glucose target and allows more aggressive correction. Using sleep mode more continuously (24/7 rather than only during scheduled sleeping hours) meaningfully shifts how the algorithm behaves overall. This is one of the primary axes on which the GNL five-level framework differentiates higher levels from lower ones.
• Basal and bolus fraction balance. In the GNL framework, the second primary axis is the balance between basal and bolus insulin share. At higher responsiveness levels, more insulin is driven through the basal channel; at lower levels, a greater share comes through bolus delivery. This reflects the different insulin delivery patterns associated with tighter versus more permissive glucose targets and correction behaviours.
• Exercise mode. A user-activated mode that raises the target glucose and suspends automatic correction boluses — useful before, during, or after planned physical activity when insulin reduction is appropriate.

2. The primary levers: Basal/bolus balance and Sleep mode

In the GNL framework, Control-IQ’s responsiveness is characterised by two interconnected levers. Understanding both is essential for interpreting the five-level structure and the explorer output it generates.

Lever 1 — Basal and bolus fraction balance

In a standard insulin regimen, there is a rough physiological expectation about how much of total daily insulin should come from basal delivery (background insulin maintaining glucose between meals) versus bolus delivery (insulin given for meals and corrections). The typical population anchor is approximately 50% basal and 50% bolus — though this varies considerably between individuals.

Control-IQ’s algorithm shifts this balance depending on how aggressively it is operating. When the system is behaving more actively — responding to predicted glucose rises with more frequent basal increases and correction boluses — a larger proportion of daily insulin tends to come through the basal channel. When operating more permissively — less frequent correction, lower insulin delivery overall — the bolus fraction typically carries a greater share.

The GNL framework formalises this into the five-level table. At level 5 (very high), the estimated basal fraction is 65% and bolus 35%. At level 1 (very low), the estimated basal fraction is 45% and bolus 55%. These are not settings entered into the pump — they are the engine’s characterisation of the expected delivery pattern at each responsiveness level.

Lever 2 — Sleep mode and its duration

Sleep mode in Control-IQ is a distinct operating state with a lower glucose target and more aggressive correction behaviour. In standard mode, the system targets a glucose range of approximately 6.7–10.0 mmol/L (120–180 mg/dL). In sleep mode, the lower target is more aggressively maintained — the algorithm aims tighter and corrects more decisively when glucose rises above threshold.

The key insight in the GNL framework is that how long sleep mode runs per day substantially changes overall algorithm behaviour. Using sleep mode only during usual sleeping hours (typically 7–8 hours overnight) produces a meaningfully different average delivery pattern than running it continuously for 24 hours a day. At GNL levels 4 and 5, sleep mode is applied 24/7 in the framework’s model. At levels 1–3, it applies only during usual sleeping hours.

Many people find that running sleep mode continuously produces tighter average glucose — particularly overnight and in the early morning — but requires more consistent carbohydrate management and carries a higher potential for nocturnal hypoglycaemia if other settings are not well-matched. This is worth exploring with your care team before making changes to your sleep mode schedule.

How this compares to the other optimisation strategy

Control-IQ and CamAPS FX share the same fundamental approach: increase algorithm-delivered insulin by pushing basal higher (up to 65% of TDD) and setting weaker carb ratios so the remaining bolus insulin is divided by actual carb intake. This gives the algorithm the most dynamic range to adapt. Both track IOB more physiologically than a simple linear decay. Control-IQ uses a fixed 5h AIT — IOB is visible but decays on a linear model that diverges from physiological insulin action.

This is fundamentally different from Medtronic 780G and Omnipod 5, which lower IOB visibility by using a shorter AIT. Those systems assume insulin clears faster, enabling more frequent auto-corrections — but at the cost of reduced IOB visibility, which creates a risk during and after exercise. The IOB physiology mismatch is explored in the IOB Guide.

3. The five levels

The GNL explorer maps Control-IQ settings onto five levels, characterised by basal fraction, bolus fraction, correction factor rule, and sleep mode pattern. The table below gives the reference values for all five levels. The descriptions that follow explain what each level means in practice — what the delivery pattern implies, what behaviour requirements it tends to demand, and when it might be most appropriate.

Level 5 — Very high responsiveness (sleep mode 24/7, basal 65%)

At level 5, the GNL framework applies sleep mode continuously — 24 hours a day, 7 days a week. This means the algorithm is consistently targeting the tighter sleep-mode glucose range and applying its more aggressive correction posture around the clock, not just overnight. The estimated basal fraction at this level is 65%, with 35% of daily insulin coming through the bolus channel.

The correction factor rule at level 5 is 80 (mmol/L) or 1400 (mg/dL) — the most aggressive correction factor in the framework, indicating the engine expects each unit of insulin to produce a smaller correction per unit than at lower levels. This reflects the typical pattern in people running very active AID delivery: the algorithm is giving more frequent, smaller increments rather than infrequent larger corrections.

Level 5 tends to be most appropriate when the priority is minimising time above range, when hypoglycaemia risk has been carefully assessed and managed, and when the user has well-dialled programmed settings — particularly carb ratios and correction factors — that are already close to their individual physiology. Running sleep mode 24/7 is a demanding posture: it requires consistent carbohydrate management and attentive use of bolus delivery at mealtimes, since the system is providing less tolerance for post-meal rises than at lower levels.

Level 4 — High responsiveness (sleep mode 24/7, basal 60%)

At level 4, sleep mode continues to run continuously, and the basal fraction is estimated at 60%, with 40% through bolus delivery. The correction factor rule is 85 (mmol/L) or 1500 (mg/dL) — slightly more permissive than level 5 but still in the more active range of the framework.

This level retains the 24/7 sleep mode pattern while easing the basal/bolus balance one step from the most aggressive position. Many people find that level 4 represents a useful balance between tight overnight and daytime control (from the continuous sleep mode) and slightly more tolerance in the bolus channel compared to level 5. It may be appropriate for experienced users who want the benefits of continuous sleep mode but with a degree of additional flexibility.

Level 3 — Medium responsiveness (sleep mode overnight, basal 55%)

Level 3 is the neutral reference point in the GNL framework. Sleep mode applies only during usual sleeping hours — the algorithm returns to standard mode during waking hours. The basal fraction is estimated at 55% and bolus at 45%. The correction factor rule is 90 (mmol/L) or 1600 (mg/dL) — the neutral midpoint of the framework’s five values.

Level 3 represents how Control-IQ typically behaves in a straightforward default configuration: sleep mode used as designed for overnight hours, and programmed settings doing most of the work during the day. This level is often an appropriate starting point and is a useful reference when interpreting the explorer output for people who are new to the system or who are not using any of the more advanced sleep mode strategies.

Level 2 — Low responsiveness (sleep mode overnight, basal 50%)

At level 2, sleep mode remains limited to usual sleeping hours, and the estimated basal fraction is 50% — with 50% of daily insulin coming through the bolus channel. The correction factor rule is 100 (mmol/L) or 1800 (mg/dL), indicating a more permissive correction expectation.

A bolus-dominant delivery pattern at level 2 tends to be associated with systems that are less aggressively correcting via basal modulation between meals. The algorithm is doing less between-meal heavy lifting, and a greater proportion of glucose management is occurring through mealtime bolusing. This profile may be relevant during periods of high physical activity (where basal reductions are common), during illness recovery, or when the priority is reducing daytime automation intensity while retaining nighttime sleep-mode protection.

Level 1 — Very low responsiveness (sleep mode overnight, basal 45%)

Level 1 is the most permissive posture in the framework. Sleep mode applies only overnight, and the estimated basal fraction is 45% — with 55% of daily insulin through bolus delivery. The correction factor rule is 110 (mmol/L) or 2000 (mg/dL), the most permissive in the framework.

This delivery profile is characterised by minimal between-meal automated correction and a heavy reliance on bolus insulin for glucose management. The system is providing less proactive intervention during the day, and the user’s own bolusing decisions carry a larger share of the overall management load. This level may be appropriate when hypoglycaemia risk is the primary concern, when activity levels are very high and unpredictable, or as a transitional starting posture when first using the system and confidence with the algorithm’s behaviour is still developing.

4. How the shared engine works

The GNL Multi-System Explorer uses a shared physiological engine to generate exploratory output ranges across all four AID systems it covers. Understanding the engine’s logic helps interpret the numbers it produces for Control-IQ.

Sensitivity class

The engine first classifies insulin sensitivity using total daily dose (TDD) divided by body weight in kilograms, expressed as units per kilogram per day (U/kg/day). This ratio is a well-established population-level proxy for overall insulin sensitivity. A lower ratio tends to indicate higher sensitivity; a higher ratio tends to indicate lower sensitivity. It anchors the correction factor and carb ratio estimates that follow.

Sensitivity classification is a starting estimate, not a fixed label. Individual sensitivity shifts with illness, activity, hormonal cycles, changes in weight or body composition, and many other factors. The sensitivity class is the engine’s reference point for generating a plausible initial range — not a characterisation of a fixed physiological trait.

Correction factor rule

The engine derives a base correction factor (CF) using a CF rule divided by TDD. This approach — a variant of the widely used “1500 rule” or “1800 rule” depending on units — estimates how many mmol/L (or mg/dL) one unit of insulin is likely to lower glucose on average. The five Control-IQ levels each apply a different CF rule value (80–110 in mmol/L terms, 1400–2000 in mg/dL terms), capturing how effective correction behaviour tends to differ across the spectrum from very active to very permissive delivery patterns.

These CF rule values are used in the engine’s calculations to generate exploratory ranges. They are not identical to the correction factor entered into the t:slim X2 pump, though they are related. The programmed correction factor in the pump is the value the user and care team set; the engine’s CF estimate is derived from population-level rules at the selected level and may differ from any individual’s programmed value.

Basal and bolus distribution across time blocks

The engine distributes estimated insulin delivery across time blocks representing morning, afternoon, evening, and overnight periods. For Control-IQ, the basal fraction at each level (45%–65% depending on level) is applied to the background delivery estimate, and the bolus fraction is applied to the meal-and-correction delivery estimate. These fractions reflect the expected delivery pattern at each level — not fixed settings programmed into the pump.

Carb ratio logic

The engine uses two carb ratio approaches. Where the user provides carbohydrate intake information, a carb-informed ratio is derived from TDD and estimated carb consumption. Where carb data is not available, a rule-based estimate anchored to TDD and sensitivity class is used instead.

For Control-IQ, the carb ratio in the engine’s output is an exploratory reference range. The carb ratio programmed into the t:slim X2 pump directly shapes how large a bolus the system delivers when a meal is announced. Because Control-IQ does not have a learning algorithm that adjusts the carb ratio over time, the accuracy of the programmed carb ratio has a direct effect on post-meal glucose excursions. Getting this value well-calibrated to individual physiology matters more for Control-IQ than for systems with stronger learning components.

5. How to use the explorer output

The AID system explorer generates a set of exploratory output ranges for Control-IQ based on the values you enter. Here is how to interpret those numbers with appropriate context.

What the output is

The numbers are directional estimates. They describe where the physiological engine calculates average responses to sit, given a particular combination of TDD, weight, and selected responsiveness level. They are generated from population-level rules and physiological anchor values — not from your individual glucose data, pump settings, or CGM history.

The output is most useful as a frame of reference. Does the correction factor the engine suggests sit in a plausible range given your experience with the system? Does the carb ratio estimate align with what you observe on your CGM after meals? When the output diverges meaningfully from what you observe in practice, that divergence is worth exploring — it may point to individual variation, or to programmed settings that need re-examination with your care team.

What the output is not

• It is not a recommendation to change your pump settings, mode, or sleep schedule.
• It is not a calculation of your personal correction factor or carb ratio for programming into the t:slim X2.
• It is not derived from your CGM data, your TIR history, or your actual delivered insulin.
• It does not reflect real-time changes in sensitivity due to illness, hormones, activity, or other factors.
• It does not account for the interaction between your specific programmed basal rates and Control-IQ’s modulation behaviour.

How to use it well

Many people find the explorer most useful when they bring the output to a clinical review. The ranges can anchor productive questions: “The engine estimates my correction factor at level 3 to be around X — does that match what my CGM shows after corrections?” That kind of question — linking the explorer output to actual glucose data — tends to generate more useful conversations than either source alone.

It is also worth exploring how the output changes as you move between levels. The shift in correction factor between level 1 and level 5 illustrates the mechanism behind why sleep mode and basal fraction matter: the same TDD produces a different average correction behaviour depending on how the insulin is distributed and how actively the algorithm is operating. The exact numbers in any individual case will differ from the engine’s estimates — but the directional relationship tends to hold.

For Control-IQ specifically, the sleep mode dimension of the level is worth exploring in its own right. Comparing level 3 (overnight sleep mode) and level 4 or 5 (sleep mode 24/7) in the explorer output can illustrate how continuous sleep mode changes the estimated delivery pattern — providing a concrete reference point for discussing this option with your care team.