Nous Research fine-tuned Mistral 7B with a custom dataset built specifically for structured tool use and function calling at small scale. Using the ChatML prompt format and a dedicated <tool_call> token, it handles nested function schemas and complex JSON output at 7B parameters, scoring competitively with models ten times its size on agentic benchmarks.
Discover the power and diversity of large language models available with Telnyx. Explore the options below to find the perfect model for your project.
| Organization | Model Name | Tasks | Languages Supported | Context Length | Parameters | Model Tier | License |
|---|---|---|---|---|---|---|---|
| No data available at this time, please try again later. |
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.
Hermes 2 Pro Mistral 7B builds on the strong Mistral 7B base with enhanced function calling and structured output capabilities. It is a capable model for tool-using applications at the 7B scale.
Hermes 2 Pro uses ChatML format for conversation templating and supports structured JSON outputs for function calling. It is available in standard and GGUF quantized formats on Hugging Face.
Hermes 2 Pro Mistral 7B scores 62.2% on MMLU, comparable to Nous Hermes 2 Mistral 7B DPO (63.4%) on the same sheet. Its differentiation is in function calling, where it scores competitively with models 10x its size through a dedicated tool_call token format and built-in JSON mode. For general knowledge it trails Gemma 7B IT (64.3%) by about 2 points.
The cost of running Hermes 2 Pro Mistral 7B with Telnyx Inference is $0.0002 per 1,000 tokens. Processing 1,000,000 function-calling tasks at 1,000 tokens each would cost $200, the same as other 7B-class models but with structured JSON output and tool-use capabilities that typically require larger models.
The main limitations are the 7B parameter count constraining complex reasoning, and a 32K context window. For tasks requiring deeper analysis, larger models in the Mixtral or Llama families are better suited.
Hermes 2 Pro adds structured function calling and JSON output mode to the base Mistral 7B, making it one of the first small models with reliable tool-use capability. It is available through Telnyx and other providers.
Yes, Hermes 2 Pro Mistral 7B is released under the Apache 2.0 license for free commercial use. Weights are on Hugging Face.
