* This blog post is a summary of this video.

Anthropic Launches GPT-4 Turbo, Custom Models, and More in OpenAI Dev Day 2024

Author: CNETTime: 2024-01-29 03:30:00

Table of Contents

Introducing GPT-4 Turbo with 128,000 Token Context Length

GPT-4 Turbo is the latest generative AI model from Anthropic. It builds on the capabilities of the previous GPT-4 model by supporting up to 128,000 tokens of context length. This expanded context allows the model to take into account much more information when generating text, leading to more accurate and relevant outputs.

The previous GPT-4 model was limited to only 8,192 tokens of context. While impressive, this could constrain the model's understanding of a prompt in cases where more context was necessary. By increasing the context length over 15 times, GPT-4 Turbo removes a major bottleneck that limited more advanced applications.

Previous Limitations of GPT-4

The previous GPT-4 model from Anthropic was groundbreaking in its capabilities, but had a context length limit of 8,192 tokens. For many use cases, this was more than enough context to produce high quality results. However, some advanced applications in areas like personalized chat, custom writing, and contextual Q&A required the model to reference much more background information in order to respond appropriately. With only 8,192 tokens of context, GPT-4 would often lose the broader context when prompted with lengthy background information. This could result in contradictory or irrelevant responses.

Expanded Context Length in GPT-4 Turbo

GPT-4 Turbo solves the context length limitation with support for up to 128,000 tokens. This allows the model to ingest entire documents, long conversations, and rich background information to truly understand the context before responding. Early benchmarks show that the additional context significantly improves accuracy and relevance across a variety of advanced use cases. By removing the context bottleneck, GPT-4 Turbo opens the door to more impactful applications of generative AI.

Updated Knowledge Cutoff to April 2023

In addition to increased context length, GPT-4 Turbo also contains updated knowledge about the world up until April 2023. The previous GPT-4 model only had knowledge until 2021, which could cause the model to be unaware of major recent events.

By updating the knowledge cutoff by over 2 years, GPT-4 Turbo can discuss and reference more current events and facts. This prevents outdated responses and opinions, improving the safe real-world use of the model.

New Text-to-Speech Capabilities

GPT-4 Turbo comes with integrated text-to-speech capabilities using advanced deep learning models. Users can now generate audio files directly from the model by providing text prompts. This allows for a wide range of new applications in accessibility, media, entertainment and more.

The text-to-speech includes 6 preset voices to choose from, with impressively human-like results. Early testers have been stunned by how natural the generated audio sounds. The voices can even convey emotion and emphasis based on the text prompt provided.

Custom Models Program for Customized AI

Alongside the launch of GPT-4 Turbo, Anthropic is introducing a new Custom Models program. This program allows companies to work directly with Anthropic's researchers to develop custom AI models fine-tuned for their specific use case.

The custom models leverage all of the capabilities of GPT-4 Turbo but are specialized even further with custom instructions, expanded niche knowledge, and custom actions. This level of customization allows companies to truly tailor conversational AI to their unique needs.

Doubling Tokens Per Minute for Established Customers

To empower companies with the compute resources necessary to take advantage of GPT-4 Turbo's capabilities, Anthropic is doubling the tokens per minute provided to established API customers. This makes it faster and cheaper to integrate GPT-4 Turbo into existing applications.

With more tokens available per minute, established customers will be able serve more of their end users with AI-generated content. The increased throughput allows companies to do more with conversational AI without being bottle-necked.

Increased Limits and Quotas in API

Alongside doubling tokens per minute, Anthropic is also increasing default rate limits and quotas across all tiers of API access. Customers will have higher baselines for requests per minute/second, allowing more users simultaneous access.

For those needing even higher volumes, custom quotas can now be requested directly through the API dashboard. This self-service model makes it faster to provision the necessary capacity for rapidly scaling applications.

Copyright Shield for Legal Protection

To provide peace of mind around legal concerns, all GPT-4 Turbo customers are covered under Anthropic's new Copyright Shield program. This program legally defends customers and covers any costs arising from copyright claims.

This protection allows companies to confidently deploy GPT-4 Turbo knowing that any rare copyright issues will be handled completely by Anthropic. With the shield in place, customers can focus entirely on leveraging conversational AI for their business rather than worrying about compliance.

Cheaper Pricing for GPT-4 Turbo Prompt and Completion Tokens

Despite having significantly more advanced capabilities, pricing for GPT-4 Turbo is actually cheaper compared to the previous GPT-4 model. Prompt tokens are 3x less expensive, while completion tokens are 2x cheaper for GPT-4 Turbo.

The discount pricing lowers the barrier for companies and developers looking to take advantage of this powerful new model. More affordable prompt and completion token pricing allows businesses of all sizes to benefit from AI that understands context and responds appropriately.

Introducing GPT Store for Sharing Custom Models

In conjunction with the Custom Models program, Anthropic is launching the GPT Store for discovering and distributing custom conversational AI models. The GPT Store allows developers to publish custom models tuned with specialized data and instructions.

Other platforms can then integrate these published models into their own products and services. The GPT Store features curated selections of top models across different industries and use cases. Through the marketplace, customized conversational AI can make an impact beyond just the model creator.


GPT-4 Turbo represents a new level of conversational AI capabilities thanks to its 15x increase in context length over the previous GPT-4 model. Additional improvements like updated knowledge, integrated text-to-speech, custom model building, cheaper pricing, and legal protection further establish GPT-4 Turbo as an extremely advanced generative AI suitable for enterprise use.

Companies across all industries now have access to customizable, contextual, and creative conversational AI to elevate customer and user experiences. GPT-4 Turbo removes previous limitations holding conversational AI back from real-world impact.


Q: What is GPT-4 Turbo?
A: GPT-4 Turbo is an upgraded version of GPT-4 that supports up to 128,000 tokens of context length, 3x more than the previous GPT-4 model.

Q: How does GPT-4 Turbo improve on GPT-4?
A: GPT-4 Turbo has a much larger context length, updated knowledge cutoff, and is more affordable due to cheaper pricing per token.

Q: What is the Custom Models program?
A: The Custom Models program allows Anthropic to work with companies to develop customized AI models tailored to their specific use case and needs.

Q: What is the GPT Store?
A: The GPT Store allows users to publish and share customized GPT models for others to use, with curation by Anthropic.

Q: How does Copyright Shield protect customers?
A: Copyright Shield protects customers by having Anthropic cover any legal costs related to copyright infringement claims when using their services.

Q: What types of voices are available for text-to-speech?
A: There are 6 natural sounding preset voices available for converting text to audio speech.

Q: How can I increase API limits and quotas?
A: API limits and quotas can be increased by directly requesting changes through your Anthropic API account settings.

Q: Will my tokens per minute be increased?
A: Yes, established GPT-4 customers will have their tokens per minute doubled for greater speed and capacity.

Q: How is GPT-4 Turbo's knowledge updated?
A: GPT-4 Turbo contains knowledge of the world up until April 2023, and will continue to be updated over time.

Q: How affordable is GPT-4 Turbo compared to GPT-4?
A: GPT-4 Turbo costs 3x less for prompt tokens and 2x less for completion tokens versus standard GPT-4 pricing.