Here are the highlights from the latest developments about GPT-4 Turbo:
GPT-4 Turbo API
- Huge 128K Context: The GPT-4 Turbo API introduces a massive 128K context, making it the largest context available in commercial Language Model (LLM) APIs. For comparison, the largest context available so far was 100K in Anthropic’s Claude 1/2.
- Affordable Pricing: GPT-4 Turbo is 4 to 6 times cheaper than GPT-4 32K. It offers pricing at just $0.01 per 1k input tokens, compared to $0.06 in GPT-4 32K. For output tokens, it costs $0.03 compared to $0.12 for GPT-4 32K.
- Enhanced Capabilities: GPT-4 Turbo with vision can now accept images along with text, and it’s available via gpt-4-vision-preview.
- Fine-Tuning Support: Developers can now fine-tune GPT-4, opening up new possibilities for customization and specialized applications.
GPT-3.5 Turbo Updates
- New API Versions: Explore the latest API versions, including gpt-3.5-turbo-1006 and gpt-4-1106-preview, with improved instruction following and other enhancements.
- Improved Function Calling: Function calling accuracy has been enhanced, allowing multiple functions to be called in a single request. The JSON mode is now available via response_format.
- Reproducible Outputs: Seed parameters are introduced to achieve reproducible outputs in the Completion API, providing more control over the model’s responses.
- Log Probabilities: Access log probabilities for the most likely output tokens in the Completion API.
Other Services and Features
- Assistants API: Build and integrate AI agents into your applications. One notable feature is the ability to create persistent and infinitely long threads, eliminating context size limits and the need to send entire conversations via Chat APIs.
- Text-to-Speech Quality: Enjoy better quality text-to-speech capabilities, enhancing the overall user experience.
- dall-e-3 Image Generator: Leveraging the power of dall-e-3, this image generator introduces new creative possibilities.
- Model Customizations for Enterprise: Tailor the models to meet the specific needs of your enterprise with model customizations.
- Whisper v3 and Consistency Decoder: Explore the latest advancements in Whisper v3 and the Consistency Decoder for improved performance and accuracy.
Prices Are Going Down
Take advantage of reduced pricing for older and new models, making AI more accessible and cost-effective:
| Older Models | New Models |
|---|---|
| GPT-4 Turbo | GPT-4 8K |
| Input: $0.03 | Input: $0.06 |
| Output: $0.06 | |
| GPT-4 32K | GPT-4 Turbo 128K |
| Input: $0.06 | Input: $0.01 |
| Output: $0.12 | Output: $0.03 |
| GPT-3.5 Turbo | GPT-3.5 Turbo 4K |
| Input: $0.0015 | Input: $0.001 |
| Output: $0.002 | Output: $0.002 |
| GPT-3.5 Turbo 16K | GPT-3.5 Turbo 4K and 16K |
| Input: $0.003 | Training: $0.008 |
| Output: $0.004 | Input: $0.003 |
| Output: $0.006 |
Stay tuned for these exciting developments in the world of AI and language models!