Built on Mistral's sparse mixture-of-experts architecture that routes each token through 2 of 8 expert networks, this Nous Research fine-tune keeps only 12.9B of its 46.7B total parameters active per forward pass. The DPO alignment stage improved its MT-Bench score over the base Mixtral Instruct while preserving the model's efficiency advantage over dense 70B-class alternatives.
Nous Hermes 2 Mixtral 8x7B DPO scores 72.3% on MMLU, slightly above the base Mixtral 8x7B Instruct (70.6%) on the same sheet after DPO alignment. Its Arena ELO of 1,084 is about 30 points below the base Mixtral Instruct (1,114), a common pattern where DPO improves knowledge scores but can shift chat preference. It matches GPT-3.5 Turbo (70.0% MMLU) at the open-weight tier.
The cost of running the model with Telnyx Inference is $0.0003 per 1,000 tokens. To put this into perspective, analyzing 1,000,000 customer chats, assuming each chat is 1,000 tokens long, would cost $300.
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 |
|---|---|---|---|---|---|---|---|
| Telnyx | DeepSeek-V4-Flash-0731 | text generation | English | 1,048,576 | 284.0B | unlisted | mit |
| Telnyx | DeepSeek-V4.1-Flash | text generation | English | 1,048,576 | 552.0B | small | mit |
| Telnyx | DeepSeek-V4.1-Flash-1017 | text generation | English | 1,048,576 | 552.0B | small | mit |
| gemma-4-26B-A4B-it | text to-text | Multilingual | 262,144 | 25.8B | small | apache-2.0 | |
| gemma-4-31B-it | text generation | Multilingual | 131,072 | 30.7B | small | apache-2.0 | |
| meta-llama | Llama-3.3-70B-Instruct | text generation | Multilingual | 99,000 | 70.6B | large | llama3.3 |
| minimaxai | MiniMax-M3-MXFP8 | text generation | Multilingual | 1,000,000 | 428.0B | small | minimax-community |
| moonshotai | Kimi-K2.6 | text generation | Multilingual | 262,144 | 1.0T | medium | modified-mit |
| moonshotai | Kimi-K3 | text generation | English | 1,000,000 | 2.8M | large | MIT Modified |
| moonshotai | Kimi-K3.1-Pro | text generation | English | 1,000,000 | 2.8B | large | kimi-k3.1-pro |
| nvidia | NVIDIA-NemotronLabs-VoiceChat-11B | speech to-speech | English | 8,192 | 11.1B | medium | openmdw-1.1 |
| openai | gpt-oss-120b | text generation | Multilingual | 131,072 | 117.0B | small | apache-2.0 |
| poolside | Laguna-S-2.1 | text generation | English | 1,048,576 | 117.6B | small | OpenMDW-1.1 |
| Telnyx | Qwen3.8-27B | text generation | English | 262,144 | 27.0B | small | Apache 2.0 |
| zai-org | GLM-5.2 | text generation | English | 1,000,000 | 753.9B | medium | MIT |
| Telnyx | GLM-5.3 | text generation | English | 1,048,576 | 753.0B | medium | glm-5.3 |
| zai-org | GLM-5.3-Flash | text generation | English | 1,048,576 | 320.0B | small | mit |
| anthropic | claude-haiku-4-5 | text generation | Multilingual | 400,000 | 0 | large | openai |
| gemini-3.7-flash | text generation | Multilingual | 400,000 | 0 | large | openai | |
| groq | gpt-oss-120b | text generation | Multilingual | 400,000 | 0 | large | openai |
| openai | gpt-4.1 | text generation | Multilingual | 400,000 | 0 | large | openai |
| openai | gpt-4o | text generation | Multilingual | 400,000 | 0 | large | openai |
| openai | gpt-4o-mini | text generation | Multilingual | 400,000 | 0 | large | openai |
| openai | gpt-5.2 | text generation | Multilingual | 400,000 | 0 | large | openai |
| openai | gpt-5.4 | text generation | Multilingual | 400,000 | 0 | large | openai |
| openai | gpt-5.4-mini | text generation | Multilingual | 400,000 | 0 | large | openai |
| openai | gpt-5.6-luna | text generation | Multilingual | 400,000 | 0 | large | openai |
| openai | gpt-5.6-sol | text generation | Multilingual | 400,000 | 0 | large | openai |
| openai | gpt-6-luna | text generation | Multilingual | 1,050,000 | 0 | large | openai |
| openai | gpt-live-1 | speech to-speech | Multilingual | 400,000 | 0 | large | openai |
Check out our helpful tools to help get you started.
Nous Hermes 2 Mixtral 8x7B DPO is a mixture-of-experts model from Nous Research, built on Mistral's Mixtral 8x7B architecture with DPO training. It features 56 billion total parameters with only a subset active per token, providing strong performance at efficient inference costs.
Nous Hermes 2 Mixtral 8x7B performs admirably on translation and complex topic understanding, and outperforms GPT-4 in certain areas like puzzles and roleplay. For general-purpose tasks, GPT-4 remains stronger overall. The model is available on multiple inference platforms as a free open-source alternative.
The model excels at multilingual reasoning, content generation, and conversational AI with a 32K context window. Its DPO training gives it strong instruction-following capabilities and improved response quality compared to the base Mixtral model.
Yes, Nous Hermes 2 Mixtral 8x7B DPO is open-source and free to use. It is available on Hugging Face and through inference providers like Ollama and OpenRouter.