Dolphin Llama 3 70B
Dolphin Llama 3 70B is an uncensored fine-tune of Meta's Llama 3 70B, developed by the Dolphin project (Eric Hartford). The Dolphin series removes the default safety refusals and RLHF guardrails from base models so that developers and researchers can build applications that require the model to engage with topics that instruction-tuned safety-aligned models decline.
At 70B scale, Dolphin Llama 3 delivers strong reasoning and long-form generation capabilities. It is intended for use by developers who need full control over content policies within their own systems rather than relying on upstream model-level filtering. Responsible use and downstream safety design remain the responsibility of the application developer.
Key Features
Uncensored instruction following without upstream refusal filters
Strong reasoning and analytical depth at 70B scale
Long-form text generation with coherent narrative flow
Developer-controlled content policy via system prompts
Multi-turn conversation with high context retention
Broad general knowledge from Llama 3 70B base
Ideal Use Cases
Research applications requiring unfiltered model responses
Custom content moderation systems built by developers
Creative writing platforms with user-controlled content settings
Security and red-teaming research environments
Applications where operator-level content governance replaces model-level filtering
Example Prompts for Dolphin Llama 3 70B
Technical Specifications
| Provider | Dolphin |
| Category | Text |
| Modality | Text -> Text |
Frequently Asked Questions
Try Dolphin Llama 3 70B now
Start using Dolphin Llama 3 70B instantly — 100 free credits, no credit card required. Access 750+ AI models through one platform.