Inception releases Mercury 3 diffusion large language model, supporting million Token context
BIBIBI
AT A GLANCE
Inception releases the diffusion large language model Mercury 3, claiming to support 1000000 Token context.
Article
Inception releases diffusion large language model Mercury 3, supporting million Token context
PANews October 20 reported, according to Inception's official announcement, that the company has released the diffusion large language model Mercury 3, supporting 1000000 Token context. The company stated that its generation speed on NVIDIA GPU reaches per second 1068 Token, with performance in multiple benchmark tests exceeding GPT-6 Luna and Claude Haiku 5.5. Standard pricing is per million input Token 0.20 USD and output Token 0.70 USD, at half price during the launch period. The model is available through Inception API, OpenRouter, Baseten, and Vercel, and is compatible with OpenAI API。
PANews reported on October 20 that Inception released the diffusion large language model Mercury 3。
02
According to Inception's official announcement, Mercury 3 supports 1000000 Token context.
03
Inception stated that the model's generation speed on NVIDIA GPU reaches per second 1068 Token。
04
Inception stated that the model outperforms GPT-6 Luna and Claude Haiku in multiple benchmark tests. 5.5。
05
Standard pricing is per million input Token 0.20 USD and per million output Token 0.70 USD, at half price during the launch period.
06
The model is available through Inception API, OpenRouter, Baseten, and Vercel, and is compatible with OpenAI API。
AI-assisted interpretation
The following is analysis, separate from reported facts. Verify important claims independently.
This is a large language model using a diffusion-based approach. The report focuses on ultra-long context, generation speed, benchmark results, usage pricing, and available access platforms; the performance and comparative results are based on Inception's official claims.
Why it matters to readers
If the relevant capabilities and interface availability are verified, developers can use it to process longer inputs and access it through multiple platforms or in a manner compatible with OpenAI API.
Beginners should focus on actual usage costs, whether the interface is easy to integrate, and whether long-context and speed performance in real-world tasks can meet the official description.
Risks and unknowns
Generation speed, 1000000 Token context capability, and benchmark results lack the independent verification indicated in the article.
The article does not provide the benchmark test names, testing methods, or detailed results.
The article does not specify the end date or specific applicable conditions for the launch half-price offer.
The article does not provide verification results for API actual availability or stability.
Related Developments
Loading event timeline…
Related concepts
Diffusion large language model
This term is not in the glossary yet. Browse related concepts in the glossary.
