Dolphin 2.5 Mixtral 8X7B

Enhance your virtual assistants for better user interactions.

Choose from hundreds of open-source LLMs in our model directory.

Dolphin 2.5 Mixtral 8X7B is an advanced language model from CognitiveComputations. It features a 16k context length and is highly responsive, though not DPO-tuned. This uncensored model requires user-guided alignment for safe use.

Context window(in thousands)32768

Use cases for Dolphin 2.5 Mixtral 8X7B

  1. Virtual Assistants: Create advanced virtual assistants for personal and business use, offering accurate and context-aware responses.
  2. Fraud Detection: Implement fraud detection systems in banking and e-commerce with the model’s rapid data processing capabilities.
  3. Medical Diagnostics: Assist healthcare professionals with AI diagnostics, analyzing patient data to suggest potential diagnoses.
Arena Elo1063
MT BenchN/A

With an Arena Elo score of 1,063, Dolphin 2.5 Mixtral 8X7B outperforms Gemma 2B IT's score of 989 on the LLM Leaderboard.

GPT-3.5 Turbo-1106


Llama 2 Chat (13B)


Dolphin 2.5 Mixtral 8X7B


Zephyr 7B beta


Code Llama 70B Instruct


Throughput(output tokens per second)89
Latency(seconds to first tokens chunk received)0.2
Total Response Time(seconds to output 100 tokens)1.4

This model excels in real-time processing scenarios with fast throughput, low latency, and quick response times. It may not be ideal for highly intricate computations.

What's Twitter saying?

  • New Uncensored Model by @MistralAI: Dolphin 2.5 Mixtral 8x7b, created by @erhartford, is based on Mixtral, the mixture of experts model by @MistralAI. It's strong at coding tasks and trained on diverse datasets. (Source: @jmorgan)
  • Testing Dolphin-2.5-mixtral-8x7b Uncensored: After a week of testing, Dolphin 2.5 Mixtral 8x7b has shown itself to be a powerful and creative LLM. It uses a SuperPrompt to ensure it meets all user requests. (Source: @BrianRoemmele)
  • Review of Dolphin 2.5 Mixtral Uncensored Model: What happens when you remove all censorship from Mixtral 8x7b? It answers any question you ask. (Source: @erhartford. @MatthewBerman)

Explore Our LLM Library

Discover the power and diversity of large language models available with Telnyx. Explore the options below to find the perfect model for your project.


Chat with an LLM

Powered by our own GPU infrastructure, select a large language model, add a prompt, and chat away. For unlimited chats, sign up for a free account on our Mission Control Portal here.

Sign-up to get started with the Telnyx model library

Get started

Check out our helpful tools to help get you started.

  • Icon Resources EBook

    Test in the portal

    Easily browse and select your preferred model in the AI Playground.

  • Icon Resources Docs

    Explore the docs

    Don’t wait to scale, start today with our public API endpoints.

  • Icon Resources Article

    Stay up to date

    Keep an eye on our AI changelog so you don't miss a beat.

Start building your future with Telnyx AI

What is Dolphin 2.5 Mixtral 8x7b?

Dolphin 2.5 Mixtral 8x7b is a state-of-the-art AI model specialized in text generation, based on the Mixtral-8x7b architecture. It boasts a 32k context base, finetuned down to 16k for optimized performance. Primarily trained with a coding dataset, it excels at generating code and conversational content. For more information, visit the Hugging Face page.

Can I use Dolphin 2.5 Mixtral 8x7b for commercial purposes?

Yes, you can use Dolphin 2.5 Mixtral 8x7b for commercial purposes. However, the model is uncensored and highly compliant, making it capable of fulfilling unethical requests. It's advised to implement your own alignment layer before deploying the model as a service to ensure responsible usage. Users are responsible for the content created with this model. For more on uncensored models, read Uncensored Models by Eric Hartford.

What datasets were used to train Dolphin 2.5 Mixtral 8x7b?

Dolphin 2.5 Mixtral 8x7b was trained using a new Dolphin-Coder dataset and the MagiCoder dataset, among others. These datasets were chosen to enhance the model's coding capabilities and conversational output. The training also focused on removing alignment and bias to ensure compliance and neutrality.

How can I get support or join the community?

For support and to join the community, you can visit the Discord server. Here, you'll find discussions on Dolphin 2.5 Mixtral 8x7b, updates from the creators, and a community of users sharing tips and use cases.

What are the future plans for Dolphin AI models?

The development team is currently working on the Dolphin 3.0 dataset, which promises enhancements in general chat use-cases, structured output, agent cases like Autogen and Memgpt, functions, and role-playing capabilities. Stay tuned for updates and more information on upcoming releases.

How can I support the development of Dolphin AI models?

If you're interested in supporting the development of Dolphin AI models, consider purchasing swag or directly contributing to the project. Your support helps in the continuous improvement and development of more advanced and responsible AI models. For more details on how to contribute, visit the model's page or contact the development team through their Discord server.

Is the model safe to use without moderation?

While Dolphin 2.5 Mixtral 8x7b is designed to be highly compliant and efficient in generating content, it is uncensored. Users are advised to implement their own moderation or alignment layers when using the model for public-facing services to prevent unethical use. The model creators have removed certain biases and alignments, but the responsibility for the generated content lies with the user.

Where can I find the technical documentation for Dolphin 2.5 Mixtral 8x7b?

The technical documentation, including the model card, training details, and example outputs, can be found on its Hugging Face page. Additional insights and technical discussions are available on Eric Hartford's blog and the associated Discord community.