What announcements are you looking forward to on the 6th November?

Personally its

  • MUCH lower cost,
  • GPT 4 fine-tuning (hopefully 32k! and GPT35 16k)
  • Higher level of reasoning for GPT4 (GPT4.5 Turbo?)
  • AGI? :smiley:

Anyway really can’t wait till the 6th!


I’m hoping for Vision API release.


Vision API, PDF upload/interpreter, Reduced pricing for gpt4, gpt4 32k general avail (w reduced pricing)


+1 for GPT 4 fine-tuning (hopefully 32k! and GPT35 16k)


I’m hoping for a context flat rate, maybe even in favor of increasing output token cost. I think this is holding the industry back, for no real reason.

But I doubt that will happen.


Vision API so I can ditch AWS Textract.


i hope it stops giving responses like this :

Below is the revised SnakeGameAI class with the necessary modifications integrated based on our discussion and the errors you encountered. Please replace the existing SnakeGameAI class with the version provided below:

Copy code

# ... rest of your methods ...

    # ... rest of your code ...
    # ... rest of your code ...

    # ... rest of your code ...
def reset(self):
    # ... (rest of your code)

def _random_food_position(self):
    # ... (rest of your code)

def _place_food(self):
    # ... (rest of your code)

def _generate_obstacles(self):
    # ... (rest of your code)

def _line_intersects_food(self, line, food):
    # ... (rest of your code)

def play_step(self, action):
# … (rest of your code)

def _move(self, action):
    # ... (rest of your code)

def _update_ui(self):
    # ... (rest of your code)

… rest of your script …

Make sure this method is included in your SnakeGameAI class, save the file, and then try running your agent.py script again. If the is_collision method is already present and correctly defined, then there might be other issues in your code causing this error. Please double-check your file and the class structure to ensure everything is in the correct place.


My guess: DALL-E 3 and GPT-4 vision, perhaps cutoff time moved further (?)

My hopes in order or priority:

  1. higher function calling limit
  2. cheaper GPT-4 32k.
  3. Lower fine-tuning cost for gpt-3.5, I probably do not care too much about finetuning GPT-4.
1 Like

I am eagerly waiting for the following to be released within this year:

  1. GPT-4 vision API
  2. DALL-E 3 API
  3. Text-to-Speech API

They have not made any major updates(?) on whisper. It’s been a year already. Perhaps they can add diarization to identify speakers in the transcript. At the very least, they should release the TTS API to complement whisper.

  1. 16k instruct model.
  2. GPT-3.5 with even larger context window. Last year Sam Altman talked about 100k and 1M context windows.
  3. Dalle3 and Vision.

Honestly, for most of my use cases gpt-3.5 is good enough, but larger context window is always welcome. Partly because that would imply the model could OUTPUT those large responses.


How OpenAI solves GPT-4’s “amnesia and hallucinations”? Because it will drive the person who is focused on discussing the problem with them crazy.


Amnesia is a consequence of token limits.
Hallucinations are the source of creativity.

I am excited for an all-in-one personalized GPT. My own personal assistant.
Vision and Dall-E API (with in/out-painting pls) would be epic.

I thought I read that enterprise accounts have their own RAG system. I would love to see that as well for API.

Lastly, tiered ChatGPT. If I can use Vision, Dall-E, and GPT-4 together I’ll be running into rate-limits way too quickly but I am more than happy to pay more.


I am looking forward to the API for voice translation and voice responses realized in ChatGPT app.


Much lower prices for at least GPT 3.5!

The current pricing is not sustainable for offering free access to the API for thousands of users via a Chrome extension.

The business model would be to provide free access with limited monthly tokens. Once users appreciate our product, they can use their own API key. It’s a good deal for us and a great deal for OpenAI with potentially thousands of new API Customers!


Open source 3.5-turbo pretrained and tuned snapshot actually from March 2023.


Absolutely, the choice to drop to summary only should be up to you the user. It’s annoying, to say the least, to have your time wasted with partial responses. Though Bard just plain blows you off and shuts down inquiry about better response with its default, “I’m just an LLM…” line, so at least you can get GPT to be more detailed if you ask it to be.

1 Like

Gpt 4.5 100k priced at 0.03
Gpt 4 32k priced at 0.01
Gpt 4 8k under 0.0080


I’d throw a lot of money away if the price was 0.03! I’m thinking 0.0003 might be better!

I guess if for 100k context it sounds worth it though I need it to be more affordable than the current price of GPT4 8k

1 Like

Welcome to the community!

Lets all work for a Better World, Brighter Future and New and Better Beginning!!!


“”" # … rest of your code …

# ... rest of your code ...

# ... rest of your code ..."""

LOL OMG this. I feel your pain it’s so real It hurts. I have begun yelling at the computer… game over :joy:

1 Like