A new model GPT-4o mini has been released. 25-30 times cheaper than GPT-4o. 82% accuracy in the reasoning and intelligence test.

Over the past month, we have added dozens of new models, such as Claude 3.5 Sonnet, NVIDIA Nemotron, Mistral Nemo, Llama 70B and more. You can select new models using the /text_model command. Our bot still has the largest collection of generative models.

For our users, the new model from OpenAI is pre-installed in the @GPT4MiniBot bot.

*?***@GPT4MiniBot

The new GPT-4o model is already in our bot.

? Works 2 times faster
? 2 times cheaper than GPT-4 Turbo
?? For Hindi it is 2.9 times cheaper
?? For Arabic it is 4 times cheaper
?? For Persian it is 3.8 times cheaper
?? For Chinese it is 2.8 times cheaper
?? For Spanish it is 2.2 times cheaper
? Works better with images.

You can select a model using the /text_model command or immediately write to the bot with a preinstalled model: @ChatGPT4oRobot

Last year we added all available neural networks to our services. You can quickly switch models using the /text_model command. This year we are updating each model on the day of its release.

Due to crazy marketing, many still think that GPT-4 is the most advanced model, but a month ago, new models from Anthropic Claude overtook the model from OpenAI in almost all benchmarks and community reviews.

The fastest model is now Meta Llama 3, Mistral and Gemma on Groq chips. These models can also be used in our service. Generating a response to the request “write an article about neural networks” takes Meta Llama 3 seconds, GPT-4 about 50 seconds with an almost identical amount of text.

The fastest model is now Meta Llama 3, Mistral and Gemma on Groq chips. These models can also be used in our service. Generating a response to the request “write an article about neural networks” takes 1 second for Mistral Groq, about 35 seconds for GPT-4 with an almost identical amount of text. And this is in Russian.

MythoMax operates without moderation.

Google Gemini has the most context.

Based on this data, it is quite obvious from the above graph that many users are switching to other models.

You can argue a lot about which model will be better, but the most correct answer will be individual for everyone. With our service you can use any model.

By the way, you can use several models at once in a group chat. This is useful for conducting tests and choosing the ideal model for you.

?@ChatGPT_AIN_bot
?‍? @Midjourney6Bot
? @Claude2NeuralBot
? @Perplexity_AI_Robot
Ⓜ️ @MetaLlamaBot
? @HeyGenRobot
? @SunoAIRobot
*?‍?***@SDXL_AIbot
in one bot: @NeuralNetworksBot

In Telegram, there are tens of thousands of neural network bots, but no one has yet fully supported GPT-4 Vision. Let's fix this!

To use the new model, send the command /text and with the following message, send one, two, or even ten images along with your request.

The GPT-4 Vision Preview model doesn't make sense to use without images, so it can't be selected in the settings. The model is automatically activated when you send a request to the bot along with an image.

If you send the command /image followed by an image with text, the request will be sent to Midjourney or Stable Diffusion, depending on your settings.

There's also a short tutorial on the video on how to generate a table through GPT and then use it with other tools.

Enjoy!

1️⃣2️⃣3️⃣4️⃣5️⃣ @MidjourneyOriginalBot

5️⃣6️⃣7️⃣8️⃣9️⃣ @ChatGPT_AIN_bot

?@ChatGPT_AIN_bot
?‍? @Midjourney6Bot
? @Claude2NeuralBot
? @Perplexity_AI_Robot
? @HeyGenRobot
? @YandexGPT_robot
? @GigaChatBySberBot
? @SunoAIRobot
*?‍?***@SDXL_AIbot
in one bot: @NeuralNetworksBot

Five hours ago, we were amazed by a fantastic presentation from OpenAI. Just five minutes ago, we received exciting news about the launch of new models. And now, literally five seconds ago, we have implemented these cutting-edge models into our bot.

Although the list of updates is extensive, I would like to highlight four key points:

1. The speed of the new models has increased several times.
2. The information has been updated to the latest data as of April 2023.
3. The cost of using the new models has been significantly reduced.
4. The amount of supported context has been expanded to 128,000 tokens.

To switch to the new models, as well as the long-available Claude, Meta Llama, Google Palm, and dozens of others, use the command /settings_text_model.

We still have the largest assortment of models for creating texts and visualizations. We are delighted that you are using cutting-edge technologies together with us.

And this is just the beginning.

1️⃣2️⃣3️⃣4️⃣5️⃣ @MidjourneyOriginalBot

5️⃣6️⃣7️⃣8️⃣9️⃣ @ChatGPT_AIN_bot

We've been a bit absent, and here's why. In August and September, we experienced incredible growth to 10 million users. This led to technical difficulties, and in September our service became unavailable twice for a short time. We offered generous compensations to everyone affected as an apology.

Throughout September, we worked diligently on improving the system. Telegram allows only 100 simultaneous requests to bots, and our bots sometimes receive significantly more requests. Now, if a bot is overloaded, it will suggest an alternative username for faster service.

We are proud of our distributed bot system, which helps to circumvent these limitations. While other developers offer paid solutions for hundreds of dollars, we provide the opportunity to create your own bot with the /newbot command for free and earn 20% from each transaction.

For those who want to create their own full-fledged solution, we offer an API with dozens of text and image generation models. To request access, write to @NeuralHelp.

And today, after OpenAI's fantastic presentation, we are excited to share with you the news about the incredible capabilities that are now available to everyone.