While GPT-4 Turbo has been around for a while in preview mode, OpenAI just released the official version.
About the Official Release
Key Highlights
- Open Access: The iconic “gpt-4-turbo” now calls upon the latest version – “gpt-4-turbo-2024-04-09”.
- Major Upgrades: Officially, the multimodal capabilities of the GPT-4 Turbo model have taken a significant leap forward.
- Image Recognition: Parsing images no longer requires an additional 4v interface.
- Context Length Extended to 128k: Still boasting an extensive contextual understanding, offering a richer interactive experience.
- Updated to Latest Data: Training data extends up to December 2023, capturing the latest pulse of the times.
For a deeper dive into interface details, visit: https://platform.openai.com/docs/models/gpt-4-turbo-and-gpt-4
Pricing
In terms of pricing, it remains consistent with the GPT-4 Turbo preview:
- Text Input: $10.00 / million tokens
- Text Output: $30.00 / million tokens
- Image Processing: Starting at just $0.00085 / image
For more pricing information, see OpenAI’s pricing policy: https://openai.com/pricing
Plus Account 40 Message Limit Every 3 Hours Removed?
According to multiple media reports, OpenAI has replaced the 40-message usage limit for GPT-4 every 3 hours with dynamically adjusted flexible quotas based on supply and demand. OpenAI will increase or decrease users’ requests to GPT-4 based on the current system’s busyness.
Official Version Frequency Limits
Looking at the top-tier Tire 5 level, the new rules are as follows:
- Maximum Concurrent Usage: 10,000 times / minute
- Maximum Processing Speed: 1,500,000 tokens / minute
For details, please refer to: https://platform.openai.com/docs/guides/rate-limits/usage-tiers?context=tier-five
Although the backend still displays the old limits, I believe the official update will be completed soon:
- Current Maximum Concurrent Usage: 3,000 times / minute
- Current Maximum Processing Speed: 250,000 tokens / minute
To check your personal limits, visit: https://platform.openai.com/account/limits
Other Versions of GPT-4
Besides today’s star, GPT-4 Turbo, let’s take a look back at other members of the GPT-4 series:
🌟 GPT-4: Released on March 14 last year (i.e., 0314 version), later updated to 0613 version, supporting 8k contexts, the classic version of GPT-4.
- Input Cost: $30.00 / million tokens
- Output Cost: $60.00 / million tokens
🌟 GPT-4-32k: Released alongside GPT-4, also with 0314 and 0613 sub-versions, supporting 32k contexts, the longest model at that time, but quite expensive.
- Input Cost: $60.00 / million tokens
- Output Cost: $120.00 / million tokens
- Invitation Only
🌟 GPT-4-Turbo-Preview: Debuted on November 6 last year, followed by an update to version 0125, supporting 128k contexts, with a significant price reduction.
- Input Cost: $10.00 / million tokens
- Output Cost: $30.00 / million tokens
🌟 GPT-4-Vision-Preview: Released simultaneously with GPT-4-Turbo Preview, with the same price, but limited image processing volume provided.
- Input Cost: $10.00 / million tokens
- Output Cost: $30.00 / million tokens
- Image processing costs vary by resolution, starting at $0.00085 / image
>> Further Reading:
8 Key Upgrades in GPT-4 Turbo Release Revealed
4 Steps to Use Free GPT-4 Turbo with Copilot
5 Steps to Access Gemini 1.5 Pro
OpenAI Unveils Voice Engine
Demystifying Red Teaming: Top 10 Questions Answered!
5-Minute Read: Global AI News Roundup