Dolphin Models by Cognitive Computations
About
Dolphin is Cognitive Computations' lineage of uncensored instruction-tuned LLMs led by Eric Hartford. The family started with Llama and Mistral bases, expanded through Mixtral, Yi, Phi, and Qwen releases, and combines open instruction data for chat, coding, math, function-calling, and agent workflows. Dolphin models are optimized for user steerability and broad task compliance while intentionally leaving alignment and policy filtering to the deployer, so production use should add external safeguards.
Current Variants
Use-when guidance is derived from seed capabilities, context, release, and replacement fields.
Use when the workload needs 128k context, 72B parameters, and tool use.
Use when the workload needs 8k context and 8B parameters.
Use when the workload needs 32k context, 7B parameters, and uncensored.
Use when the workload needs 8k context, 7B parameters, and uncensored.
Use when the workload needs 8k context and 8B parameters.
Use when the workload needs 8k context, 7B parameters, and uncensored.
Use when the workload needs 128k context and 7B parameters.
Use when the workload needs 4k context and 9B parameters.
Use when the workload needs 128k context and 14B parameters.
| Model | Use when | Released | Signals | Status |
|---|---|---|---|---|
| Dolphin 2.9.2 Qwen2-72B | Use when the workload needs 128k context, 72B parameters, and tool use. | 2024-05 | 128k context72B parameterstool use | Current |
| Dolphin 2.6 Mixtral 8x7B | Use when the workload needs 32k context. | 2023-12 | 32k context | Current |
| Dolphin 2.5 Mixtral 8x7B | Use when the workload needs 32k context. | 2023-12 | 32k context | Current |
| Dolphin 2.9 Llama 3 8B | Use when the workload needs 8k context and 8B parameters. | 2023-12 | 8k context8B parameters | Current |
| Dolphin 2.7 Mixtral 8x7B | Use when the workload needs 32k context. | 2023-12 | 32k context | Current |
| Dolphin 2.9.3 Mistral 7B | Use when the workload needs 32k context, 7B parameters, and uncensored. | 2023-12 | 32k context7B parametersuncensored | Current |
| Dolphin 2.1 Mistral 7B | Use when the workload needs 8k context, 7B parameters, and uncensored. | 2023-12 | 8k context7B parametersuncensored | Current |
| Dolphin 2.9.1 Llama 3 8B | Use when the workload needs 8k context and 8B parameters. | 2023-12 | 8k context8B parameters | Current |
| Dolphin 2.2.1 Mistral 7B | Use when the workload needs 8k context, 7B parameters, and uncensored. | 2023-12 | 8k context7B parametersuncensored | Current |
| Dolphin 2.9.2 Qwen2-7B | Use when the workload needs 128k context and 7B parameters. | 2023-12 | 128k context7B parameters | Current |
| Dolphin 2.9.1 Yi1.5 9B | Use when the workload needs 4k context and 9B parameters. | 2023-12 | 4k context9B parameters | Current |
| Dolphin 2.9.2 Phi-3 Medium | Use when the workload needs 128k context and 14B parameters. | 2023-12 | 128k context14B parameters | Current |
Release Timeline
2 release groupsSpecifications(12 models)
| Model | Released | Context | Parameters | Fn Calling | Tool Use |
|---|---|---|---|---|---|
| Dolphin 2.9.2 Qwen2-72B | 2024-05 | 128k | 72B | Yes | Yes |
| Dolphin 2.6 Mixtral 8x7B | 2023-12 | 32k | 8x7B | No | No |
| Dolphin 2.5 Mixtral 8x7B | 2023-12 | 32k | 8x7B | No | No |
| Dolphin 2.9 Llama 3 8B | 2023-12 | 8k | 8B | No | No |
| Dolphin 2.7 Mixtral 8x7B | 2023-12 | 32k | 8x7B | No | No |
| Dolphin 2.9.3 Mistral 7B | 2023-12 | 32k | 7B | No | No |
| Dolphin 2.1 Mistral 7B | 2023-12 | 8k | 7B | No | No |
| Dolphin 2.9.1 Llama 3 8B | 2023-12 | 8k | 8B | No | No |
| Dolphin 2.2.1 Mistral 7B | 2023-12 | 8k | 7B | No | No |
| Dolphin 2.9.2 Qwen2-7B | 2023-12 | 128k | 7B | No | No |
| Dolphin 2.9.1 Yi1.5 9B | 2023-12 | 4k | 9B | No | No |
| Dolphin 2.9.2 Phi-3 Medium | 2023-12 | 128k | 14B | No | No |
Available From(5 providers)
Pricing
| Model | Provider | Input / 1M | Output / 1M | Type |
|---|---|---|---|---|
| Dolphin 2.6 Mixtral 8x7B | DeepInfra | $0.15 | $0.45 | Serverless |
| Dolphin 2.6 Mixtral 8x7B | Lepton AI API | $0.3 | $0.3 | Serverless |
| Dolphin 2.9 Llama 3 8B | Microsoft Foundry | $0.37 | $1.1 | Provisioned |
| Dolphin 2.6 Mixtral 8x7B | Fireworks AI | $0.5 | $0.5 | Provisioned |
| Dolphin 2.5 Mixtral 8x7B | Together AI | $0.6 | $0.6 | Serverless |
| Dolphin 2.9.2 Qwen2-72B | Fireworks AI | $0.9 | $0.9 | Provisioned |
Frequently Asked Questions
- What is Dolphin used for?
- Dolphin is used for uncensored, agent workflows and tool use, and coding. The family description and listed model capabilities point to those workloads as the best fit.
- How does Dolphin compare to Claude 3?
- Dolphin by Cognitive Computations is strongest where you need uncensored, while Claude 3 by Anthropic is the closest related family to check for vision and multimodal work. Dolphin has 12 listed variants and reaches up to 128k context, while Claude 3 reaches up to 200k context, so compare the specs and pricing tables before choosing a production model.
- Which Dolphin model should I use?
- For the lowest listed input price, start with Dolphin 2.6 Mixtral 8x7B through DeepInfra at $0.15/1M input tokens. For the most capable/latest local choice, evaluate Dolphin 2.9.2 Qwen2-72B with 128k context and tool use and function calling.
Models(12)
Dolphin 2.9.2 Qwen2-72B
Dolphin 2.6 Mixtral 8x7B
Dolphin 2.5 Mixtral 8x7B
Dolphin 2.9 Llama 3 8B
Dolphin 2.7 Mixtral 8x7B
Dolphin 2.9.3 Mistral 7B
Dolphin 2.1 Mistral 7B
Dolphin 2.9.1 Llama 3 8B
Dolphin 2.2.1 Mistral 7B
Dolphin 2.9.2 Qwen2-7B
Dolphin 2.9.1 Yi1.5 9B
Dolphin 2.9.2 Phi-3 Medium