12:33Machine-translated from Azerbaijani by AIRead in Azerbaijani
OpenAI and Anthropic Unveil Faster, More Affordable AI Models
OpenAI and Anthropic unveiled faster, more affordable artificial intelligence models on Tuesday. OpenAI’s new models are called GPT-6 Sol and GPT-6 Luna, while Anthropic’s model is called Claude Opus 5.5. According to the companies, these models deliver results close to or better than previous versions at a lower computing cost. The new models are primarily intended for developers and businesses.
GPT-6 Sol and GPT-6 Luna were developed using methods similar to those used for GPT-6 Astra, which was released in early September. The models have been improved for writing code, operating computers, and performing professional tasks. However, Astra remains the most powerful option. According to OpenAI, Sol makes approximately 2 times fewer factual errors than the previous model. On coding tasks, it reaches the level of Fable 5.1 at a lower cost. Improved memory speeds up responses by taking more information from previous conversations and queries into account. Prices are up to 50% lower than the promotional rates for the GPT-5.6 models released in July. Input tokens for GPT-6 Sol cost $2 (approximately 3 manats). Output tokens are priced at $10 (approximately 17 manats). With Luna, input tokens cost $0.10 (approximately 0 manats), while output tokens cost $0.50 (approximately 1 manat). A token is a small piece of text that an artificial intelligence model processes. Both models are becoming available in ChatGPT Work and Codex for Plus, Pro, Business, and Enterprise customers. Free users and Go subscribers will only be able to use Luna. Claude Opus 5.5 was developed for complex tasks, identifying and fixing inefficient parts of software, financial analysis, and business work. According to Anthropic’s tests, the model outperformed GPT-6 Astra in the Terminal-Bench 4.0 and FrontierCode v1.1 benchmarks, which measure autonomous coding. The company says the model’s writing and instructions are clearer. Compared with previous versions, the model made 85% fewer attempts to cross boundaries. Under standard settings, the new model costs 40% less than Opus 5 during routine work. It also requires less computing power. Input tokens for Opus 5.5 cost $4 (approximately 7 manats). Output tokens are priced at $20 (approximately 34 manats). With Opus 5, input tokens cost $5 (approximately 9 manats), while output tokens cost $25 (approximately 43 manats). Anthropic is increasing the 5-hour usage limits on its Pro, Max, and Team plans. For high-risk cybersecurity and scientific tasks, the system may switch to an older model. Developers can now use Claude Opus 5.5 on the Claude, Amazon Web Services, Google Cloud, and Microsoft Azure platforms. According to OpenAI, the new models performed better than their predecessors in safety tests. They provide less false information about the results of coding tasks. They also reject more malicious prompts that attempt to bypass safety restrictions. Both companies have also released benchmarks for evaluating artificial intelligence models that could become a common industry standard in the future.
GPT-6 Sol and GPT-6 Luna were developed using methods similar to those used for GPT-6 Astra, which was released in early September. The models have been improved for writing code, operating computers, and performing professional tasks. However, Astra remains the most powerful option. According to OpenAI, Sol makes approximately 2 times fewer factual errors than the previous model. On coding tasks, it reaches the level of Fable 5.1 at a lower cost. Improved memory speeds up responses by taking more information from previous conversations and queries into account. Prices are up to 50% lower than the promotional rates for the GPT-5.6 models released in July. Input tokens for GPT-6 Sol cost $2 (approximately 3 manats). Output tokens are priced at $10 (approximately 17 manats). With Luna, input tokens cost $0.10 (approximately 0 manats), while output tokens cost $0.50 (approximately 1 manat). A token is a small piece of text that an artificial intelligence model processes. Both models are becoming available in ChatGPT Work and Codex for Plus, Pro, Business, and Enterprise customers. Free users and Go subscribers will only be able to use Luna. Claude Opus 5.5 was developed for complex tasks, identifying and fixing inefficient parts of software, financial analysis, and business work. According to Anthropic’s tests, the model outperformed GPT-6 Astra in the Terminal-Bench 4.0 and FrontierCode v1.1 benchmarks, which measure autonomous coding. The company says the model’s writing and instructions are clearer. Compared with previous versions, the model made 85% fewer attempts to cross boundaries. Under standard settings, the new model costs 40% less than Opus 5 during routine work. It also requires less computing power. Input tokens for Opus 5.5 cost $4 (approximately 7 manats). Output tokens are priced at $20 (approximately 34 manats). With Opus 5, input tokens cost $5 (approximately 9 manats), while output tokens cost $25 (approximately 43 manats). Anthropic is increasing the 5-hour usage limits on its Pro, Max, and Team plans. For high-risk cybersecurity and scientific tasks, the system may switch to an older model. Developers can now use Claude Opus 5.5 on the Claude, Amazon Web Services, Google Cloud, and Microsoft Azure platforms. According to OpenAI, the new models performed better than their predecessors in safety tests. They provide less false information about the results of coding tasks. They also reject more malicious prompts that attempt to bypass safety restrictions. Both companies have also released benchmarks for evaluating artificial intelligence models that could become a common industry standard in the future.