[pcbeingused333]

The prompt-chatgpt function templates call gpt-3.5-turbo, which OpenAI shut down on 2026-10-23, so a freshly deployed python, node or php template answers every request with "Failed to query model."

Fix:

• use gpt-5.6-terra, the replacement OpenAI lists for gpt-3.5-turbo (https://developers.openai.com/api/docs/deprecations), in the three main files and their READMEs;
• send the token limit as max_completion_tokens instead of max_tokens, since the gpt-5 family rejects max_tokens. The OPENAI_MAX_TOKENS / MAX_TOKENS env var names are unchanged;
• drop the README sentence claiming a 4096-token limit for GPT-3.5-turbo.

Not changed: the dart template. dart_openai 3.x, the newest line compatible with its Dart 2 SDK constraint, still sends max_tokens, so fixing it needs a dependency/SDK bump that seems better as its own change. It keeps pointing at the retired model for now.

Testing: the repo has no tests for these templates, so the python (openai<0.28) and node (openai@^4.47) handlers were run against a local stub of /v1/chat/completions that answers 400 for a model outside gpt-5* or when max_tokens is in the body. On main both return 500; with this change both return 200. The php template was changed the same way but not executed. Not run against a live OpenAI key, so whether a small max_completion_tokens budget (default 512) is enough for a reasoning model's output is untested.

---

Disclosure: prepared with the assistance of an AI coding tool (Claude), submitted for human review.

[ghost]

🔵 Tier A · Mergeable after minor fixes

The model and request parameter changes are valid, but the PHP default completion budget can leave multi-step prompts with truncated or empty visible answers.

The Python, Node, and PHP prompt-chatgpt templates now use GPT-5.6-terra and send their configured limits as max_completion_tokens. Their READMEs describe the new model and remove the GPT-3.5-specific token-limit text.

Verdict	New comments	Fixed	Still open
💬 Commented	1	0	0

Note

Not approving while a bug finding is open: 150-token cap also limits GPT-5.6 reasoning

	Finding	Where
🟡	150-token cap also limits GPT-5.6 reasoning	php/prompt-chatgpt/src/index.php:26

Fix with agent prompt

### Issue 1
php/prompt-chatgpt/src/index.php:26
**150-token cap also limits GPT-5.6 reasoning**

`max_completion_tokens` counts internal reasoning tokens as well as visible answer text, so the PHP template's default budget of 150 can be consumed before a multi-step prompt gets a useful answer; the handler still returns this as a successful completion. The old GPT-3.5 `max_tokens` limit did not have this reasoning-token cost, so consider raising the default or documenting that the configured budget includes reasoning.

---

For each issue above, determine whether it is valid and should be fixed. If so, fix it directly.

📂 Walkthrough · 6

File	Change
node/prompt-chatgpt/README.md	Updates the model description and removes the GPT-3.5 token-limit statement.
node/prompt-chatgpt/src/main.js	Switches the Chat Completions request to GPT-5.6-terra and max_completion_tokens.
php/prompt-chatgpt/README.md	Updates the model description and removes the GPT-3.5 token-limit statement.
php/prompt-chatgpt/src/index.php	Switches the Chat Completions request to GPT-5.6-terra and max_completion_tokens.
python/prompt_chatgpt/README.md	Updates the model description and removes the GPT-3.5 token-limit statement.
python/prompt_chatgpt/src/main.py	Switches the Chat Completions request to GPT-5.6-terra and max_completion_tokens.

---

_{Reviewed 4e69811 · Details · Comment @hansi-codes review to re-run, or mention @hansi-codes with a question.}

[ghost]

🔵 Tier A · See the inline comments. Summary

[ghost]

150-token cap also limits GPT-5.6 reasoning

max_completion_tokens counts internal reasoning tokens as well as visible answer text, so the PHP template's default budget of 150 can be consumed before a multi-step prompt gets a useful answer; the handler still returns this as a successful completion. The old GPT-3.5 max_tokens limit did not have this reasoning-token cost, so consider raising the default or documenting that the configured budget includes reasoning.

Prompt To Fix With AI

This is a comment left during a code review.
Path: php/prompt-chatgpt/src/index.php
Line: 26

Comment:
**150-token cap also limits GPT-5.6 reasoning**

`max_completion_tokens` counts internal reasoning tokens as well as visible answer text, so the PHP template's default budget of 150 can be consumed before a multi-step prompt gets a useful answer; the handler still returns this as a successful completion. The old GPT-3.5 `max_tokens` limit did not have this reasoning-token cost, so consider raising the default or documenting that the configured budget includes reasoning.

---

For each issue above, determine whether it is valid and should be fixed. If so, fix it directly.

_{🟡 Minor · bug · Reply if this doesn't apply.}