GPT-5.4 Fast mode often feels no faster than Standard, but still consumes credits at 2x

[GGBondBlueWhale]

What version of the Codex App are you using (From “About Codex” dialog)?

Version 26.415.40636 (1799)

What subscription do you have?

Pro

What platform is your computer?

macOS

What issue are you seeing?

I’m using Codex with GPT-5.4 and I have Fast mode turned on, but in practice it often feels very slow and sometimes not meaningfully faster than Standard at all.

The current UI says Fast mode is about 1.5x faster while consuming credits at 2x, but my real-world experience lately is that the speed difference is often negligible. That makes the pricing feel unfair during periods of apparent server congestion or degraded Fast-mode capacity.

Screenshot showing Fast mode selected:

What steps can reproduce the bug?

1. Open Codex App.
2. Select GPT-5.4.
3. Enable Fast mode from the speed menu.
4. Use Codex normally across multiple prompts/tasks.
5. Compare the perceived latency with Standard mode.

In my case, Fast mode often does not feel close to the advertised improvement, while still charging at the higher 2x usage rate.

What is the expected behavior?

If Fast mode is charging 2x credits, it should consistently provide a clearly noticeable speed benefit over Standard.

If the backend is congested and Fast mode cannot provide something close to its normal performance, then I think one of these should happen instead:

• temporarily do not charge the full 2x rate, or
• show a clear indicator that Fast capacity is currently degraded, or
• automatically fall back to a lower billing multiplier when the speed benefit is not actually being delivered.

Additional information

I also want to ask whether other Pro users are seeing the same thing recently.

This seems related to Fast-mode billing / visibility discussions, but my main complaint here is specifically about the combination of:

• very little real speed improvement,
• GPT-5.4 feeling slow even with Fast enabled, and
• still being charged at the higher 2x Fast-mode rate.

From a user perspective, paying 2x while not actually receiving the expected speedup feels pretty bad.

I also believe OpenAI has a responsibility to refund all Codex usage consumed by Pro and Plus users during periods when Fast mode performance is abnormally degraded.

[Mahkhmood9]

@GGBondBlueWhale on the bright side , its still a bit faster

@etraut-openai is right about not closing this issue

maybe the Pro plan should gain a different pool list

---

on the other hand see what openai faces recently , massive influx of new users / a lot of people check it out now
2. the user complains of the 2x rates ,, and ask by not keep a continuing progress an chart based on the 0,4~ speed diffrences rate

1x = normal (baseline) speed → you pay the normal rate (let's call it 1x cost).
The 2x pool is the maximum speed boost allowed (you can't go faster than 2x).
If you choose a speed of 1.2x, you get charged 0.2x extra (i.e. total 1.2x the normal cost).

[Mahkhmood9]

The pricing formula for dynamic speed:

[
C(s) = C_0 \times s \quad \text{for} \quad 1 \leq s \leq 2
]

where ( s ) is the chosen speed multiplier and ( C_0 ) is the baseline cost at 1x speed.

Extra charge:
[
\text{Extra} = C_0 \times (s - 1)
]

def calculate_cost(s: float, C0: float = 1.0) -> dict:
    """
    Calculate cost based on chosen speed multiplier.
    
    Parameters:
        s  : Chosen speed multiplier (must be between 1.0 and 2.0)
        C0 : Baseline cost at 1x speed (default = 1.0)
    
    Returns:
        Dictionary with cost details
    """
    if not 1.0 <= s <= 2.0:
        raise ValueError("Speed multiplier 's' must be between 1.0 and 2.0")
    
    total_cost = C0 * s
    extra_charge = C0 * (s - 1)
    
    return {
        "chosen_speed": s,
        "baseline_cost": C0,
        "total_cost": round(total_cost, 4),
        "extra_charge": round(extra_charge, 4),
        "speed_description": f"{s}x speed"
    }


# Example usage:
if __name__ == "__main__":
    print("Dynamic Speed Pricing Examples:\n")
    
    for speed in [1.0, 1.2, 1.5, 1.8, 2.0]:
        result = calculate_cost(speed, C0=10.0)   # assume baseline cost is $10
        print(f"Speed: {result['chosen_speed']}x")
        print(f"  Total Cost   : ${result['total_cost']}")
        print(f"  Extra Charge : ${result['extra_charge']}")
        print(f"  Baseline     : ${result['baseline_cost']}\n")

[etraut-openai]

@Mahkhmood9, please stop spamming our issue tracker.

[damonpkuml]

I’ve had the same experience — Fast Mode only increases token usage.