New version consuming lot more tokens

Codex
1

After updating from snap to the official version, codex started consuming lot more tokens.

javier@jm:~/proyectos/emprendimientos/lonca/shipii/src/Shipii$ sudo snap refresh codex --edge
codex --version
codex (edge) 0.146.0 desde jcat (jcat-nysasounds) refreshed
codex-cli 0.146.0
javier@jm:~/proyectos/emprendimientos/lonca/shipii/src/Shipii$ codex --version
codex-cli 0.146.0
javier@jm:~/proyectos/emprendimientos/lonca/shipii/src/Shipii$ codex exec "respondé únicamente OK"
OpenAI Codex v0.146.0
--------
workdir: /home/javier/proyectos/emprendimientos/lonca/shipii/src/Shipii
model: gpt-5.4
provider: openai
approval: never
sandbox: workspace-write [workdir, /tmp, $TMPDIR]
reasoning effort: none
reasoning summaries: none
session id: 01a00258-638f-7492-9bfa-94e43a1c63be
--------
user
respondé únicamente OK
warning: Codex could not find bubblewrap on PATH. Install bubblewrap with your OS package manager. See the sandbox prerequisites: . Codex will use the bundled bubblewrap in the meantime.
codex
OK
tokens used
9.645
javier@jm:~/proyectos/emprendimientos/lonca/shipii/src/Shipii$ $HOME/.local/bin/codex exec "respondé únicamente OK"
OpenAI Codex v0.147.0
--------
workdir: /home/javier/proyectos/emprendimientos/lonca/shipii/src/Shipii
model: gpt-5.6-sol
provider: openai
approval: never
sandbox: workspace-write [workdir, /tmp, $TMPDIR]
reasoning effort: none
reasoning summaries: none
session id: 01a00268-5001-7970-8827-3ca464ae3e68
--------
user
respondé únicamente OK
codex
OK
tokens used
10.944
javier@jm:~/proyectos/emprendimientos/lonca/shipii/src/Shipii$ $HOME/.local/bin/codex exec "respondé únicamente OK" \
  2>&1 | tee /tmp/codex_standalone_test.log

grep -E \
  'unknown variant|failed to refresh available models|models_manager|bubblewrap' \
  /tmp/codex_standalone_test.log
OpenAI Codex v0.147.0
--------
workdir: /home/javier/proyectos/emprendimientos/lonca/shipii/src/Shipii
model: gpt-5.6-sol
provider: openai
approval: never
sandbox: workspace-write [workdir, /tmp, $TMPDIR]
reasoning effort: none
reasoning summaries: none
session id: 01a00269-2aa5-7ea2-8ad5-a6b60ef79fc0
--------
user
respondé únicamente OK
codex
OK
tokens used
10.944
OK
javier@jm:~/proyectos/emprendimientos/lonca/shipii/src/Shipii$ sudo snap remove codex
hash -r
codex eliminada (instantánea de datos de snap guardada)
javier@jm:~/proyectos/emprendimientos/lonca/shipii/src/Shipii$ type -a codex
command -v codex
readlink -f "$(command -v codex)"
codex --version
codex is /home/javier/.local/bin/codex
/home/javier/.local/bin/codex
/home/javier/.codex/packages/standalone/releases/0.147.0-x86_64-unknown-linux-musl/bin/codex
codex-cli 0.147.0
javier@jm:~/proyectos/emprendimientos/lonca/shipii/src/Shipii$ codex exec -m gpt-5.4 "respondé únicamente OK"
OpenAI Codex v0.147.0
--------
workdir: /home/javier/proyectos/emprendimientos/lonca/shipii/src/Shipii
model: gpt-5.4
provider: openai
approval: never
sandbox: workspace-write [workdir, /tmp, $TMPDIR]
reasoning effort: none
reasoning summaries: none
session id: 01a0026e-763e-7be2-a889-8d3ff182cc3b
--------
user
respondé únicamente OK
codex
OK
tokens used
14.245
javier@jm:~/proyectos/emprendimientos/lonca/shipii/src/Shipii$ codex exec --json -m gpt-5.4 \
  "respondé únicamente OK" \
  | tee /tmp/codex_gpt54.jsonl
{"type":"thread.started","thread_id":"01a00270-49ee-7e21-b777-5363ea82b01f"}
{"type":"turn.started"}
{"type":"item.completed","item":{"id":"item_0","type":"agent_message","text":"OK"}}
{"type":"turn.completed","usage":{"input_tokens":19719,"cached_input_tokens":5504,"cache_write_input_tokens":0,"output_tokens":19,"reasoning_output_tokens":12}}
javier@jm:~/proyectos/emprendimientos/lonca/shipii/src/Shipii$ jq 'select(.type == "turn.completed") | .usage' \
  /tmp/codex_gpt54.jsonl
{
  "input_tokens": 19719,
  "cached_input_tokens": 5504,
  "cache_write_input_tokens": 0,
  "output_tokens": 19,
  "reasoning_output_tokens": 12
}
javier@jm:~/proyectos/emprendimientos/lonca/shipii/src/Shipii$
2

Look, I used maybe 3–4 commands and I’ve already hit the weekly limit. This is honestly getting ridiculous — I can’t get any work done like this. I can’t afford to keep paying more and more, and my quota seems to run out faster every time…

FOR THE LOVE OF GOD, PLEASE GIVE US A HIGHER LIMIT!

3

It is not a matter of give a higher limit, is that the model is supposedly consuming lots of credits but they never say how they are uses. I feel scammed by openai. As I have already shown, near 20 thousand credits were used to execute the prompt “reply only OK”, which is absurd. We can fill the openai support form or ask god to solve it (with almost same result in both actions), but codex should be clear on what is consuming.