ChatML prompt format
According to the model card, the model was trained using the ChatML prompt template format with documented example turns.
Open Source Model Profile · dphn
dolphin-2.9-llama3-8b
dolphin-2.9-llama3-8b is an 8.03B-parameter Llama text-generation fine-tune from dphn. According to the model card, it is based on Llama-3-8B with ChatML formatting, function-calling support, and uncensored, compliance-oriented tuning.
- Publisher
- dphn
- Task
- text-generation
- Model type
- llama
- License
- other
- Library
- transformers
- Publication status
- Accepted · not indexed
Model overview
dolphin-2.9-llama3-8b is published by dphn as a Llama text-generation model in the Dolphin 2.9 series. The captured configuration identifies LlamaForCausalLM and Safetensors metadata reports 8,030,277,632 parameters. According to the model card, it is a full-weight fine-tune of Llama-3-8B with a 4K sequence length, ChatML prompt template, and agentic function-calling abilities.
Recorded capabilities
Function calling
According to the model card, Dolphin-2.9 supports function calling alongside instruction, conversational, and coding skills with initial agentic abilities.
Uncensored tuning posture
According to the model card, the dataset was filtered to remove alignment and bias, making the model highly compliant, and the publisher advises adding an alignment layer before serving it.
Documented training run
According to the model card, full-weight fine-tuning took 2.5 days on 8x L40S GPUs with axolotl 0.4.0, three epochs, and a 2e-5 learning rate.
Use cases in the source record
- Instruction, conversational, and coding text generation using the documented ChatML prompt template.
- Function-calling and agentic experiments using the initial agentic abilities the model card describes.
Limitations and unknowns
- The captured record lists license other; the model card states the model is governed by the Meta Llama 3 Community License Agreement, so redistribution terms need confirmation.
- The model card reports a dataset bug causing over-mention of the system message; the publisher recommends directing the model not to mention it unless asked.
- Publisher-reported training-loss and validation-loss figures are source claims and were not independently verified.
- Provider state is historical snapshot data, not independently refreshed current availability.
Source and provenance
Source: dphn/dolphin-2.9-llama3-8b
Captured: Unknown. Processed: 2026-09-07T19:34:43.834131+00:00.
Dolphin 2.9 Llama 3 8b 🐬 Curated and trained by Eric Hartford, Lucas Atkins, and Fernando Fernandes, and Cognitive Computations A bug has been found in the Dolphin 2.9 dataset in SystemConversations that causes the model to overly talk about the "SYSTEM MESSAGE". To counter this, we recommend you add a statement in the system message directing the model not to mention the system message. An example system message is "The assistant is named Dolphin. A helpful and friendly AI assistant, Dolphin avoids discussing the system message unless directly asked about it." My appreciation for the sponsors of Dolphin 2.9: Crusoe Cloud - provide…
F001F002F003F004F005F006F007F008F009F010F011F012F013F014F015F016F017F018F019F021F022