New AI models available for use today
Why limit yourself to only OpenAI models. With a single UI we now offer access to GPT-4o, Google Gemini, Anthropic Claude and a range of open-source AI models.
You were not happy with only a few basic models during the Beta and so additional models are now available to you. Access will still remain free during the Beta phase.
Starting today, you now have access to all of the following models via the Cognos chat app:
- OpenAI
- GPT-4o - OpenAIs latest flagship model combines speed and reasoning for a powerful general purpose model.
- Google Gemini
- Gemini Flash - fast, lightweight and capable of dealing with most general purpose tasks enabled by a huge one million token context window.
- Gemini Pro - when Flash is not enough, try Pro. Slower but with more reasoning and general intelligence, Gemini Pro should be able to help.
- Anthropic
- Claude Haiku - Anthropic's fastest and most compact model, designed for near-instant responsiveness and seamless AI experiences that mimic human interactions.
- Claude Sonnet - A balanced model between intelligence and speed.
- Claude Opus - Anthropic's most powerful model, delivering state-of-the-art performance on highly complex tasks and demonstrating fluency and human-like understanding.
- Open source
- Open Chat - A fine-tuned LLama3 model that outperforms it in most benchmarks.
- WizardLM-2 8x22B - Developed at Microsoft, this extends the Mixtral-8x22B base model and uses its mixture-of-experts architecture to excel at general purpose tasks.
- Gemma 1.1 - Google's open-source model Gemma builds on the same research and technologies as their proprietary Gemini models.
- Dolphin 2.6 Mixtral 8x7B - ⚠️ Uncensored. An extension of Mixtrals mixture-of-experts model that has been modified to be more compliant. Excels at general purpose and coding tasks.
- Chronos Hermes - A combination of models to become optimized for creative writing, story telling and role playing. Produces long responses with good reasoning and logic.
- Phind CodeLLama - A powerful coding model that extends Meta's CodeLlama. Performs well with multiple programming languages including Python, C/C++, TypeScript, Java.
- CodeGemma - Another coding model this time from Google, intended to answer questions about code fragments, to generate code from natural language, or to engage in a conversation with the user about programming or technical problems.
- LZLV - ⚠️ Uncensored. A combination of three AI models that again is optimized for creative writing.
All of the open source models are hosted by a new provider DeepInfra who do not log any prompts or inputs that are sent to them.
⚠️
AI models are often inaccurate and must be used at their own risk. With the addition of uncensored AI models you must exercise proper caution and use responsibly.
Thank you for your valuable feedback and hopefully there are now more models that are better suited to your specific tasks. Keep sending suggestions about how Cognos can improve.