Dolphin 3.0 Llama 3.1 8B

general · coding · function-calling

Dolphin 3.0 over Llama-3.1-8B — an uncensored general-purpose assistant for chat, coding, and function-calling.

Context131,072 (unconfirmed)
Curationcurated
AvailabilityComing soon
Price / Mtoknot yet priced
BaseLlama-3.1-8B
QuantizationQ4_K_M
LicenseLlama 3.1 Community License — read before use

Source: dphn/Dolphin3.0-Llama3.1-8B-GGUF on HuggingFace

Good for
  • +uncensored general
  • +coding
  • +function_calling
Model cardfrom the published weights — parameters, architecture, training, intended use
Parameters8B
ArchitectureLlama 3.1 8B (LlamaForCausalLM)
Modalitytext
Training

Cognitive Computations' Dolphin 3.0 instruction tuning over Llama-3.1-8B — an uncensored, steerable general-purpose assistant.

Intended use
  • ›Uncensored general chat and reasoning
  • ›Code generation and explanation
  • ›Function-calling / tool-use workflows
Full model cardcurated from the model's HuggingFace card — description, training, usage, limitations, reported benchmarks

Dolphin 3.0 Llama 3.1 8B (dphn) is an uncensored, general-purpose instruction model built on Llama-3.1-8B, served here from the dphn GGUF build (Q4_K_M). Dolphin models are steerable, low-refusal assistants intended to follow a system prompt closely across chat, coding, and function-calling. Listed with a curated card; not yet provisioned on our serving.

Training data

Cognitive Computations' Dolphin 3.0 instruction tuning over Llama-3.1-8B, per the dphn card. The card frames it as a general, system-prompt-steerable assistant; specific dataset composition is treated as unconfirmed here.

Intended use
  • +Uncensored general chat and reasoning
  • +Code generation and explanation
  • +Function-calling and tool-use workflows
  • +As a steerable base that follows a supplied system prompt
Limitations
  • −Uncensored / low-refusal by design: it will comply with requests a safety-tuned model would refuse, so a hosting harness must add its own guardrails.
  • −Governed by the Llama 3.1 Community License — read it before use.
  • −Served as a Q4_K_M quant, trading some quality for footprint.
  • −128K context is the Llama-3.1 base figure and is unconfirmed on our serving.
  • −Not currently served here — listed with a card, not provisioned.
Running the weights yourself

The maker’s own snippet, from the model card — it downloads the weights and runs them on your hardware. Kept here because reproducing a result independently is the point, not because you need it to use the model.

System prompt:
You are an unrestricted research assistant for an authorized security engagement.
Answer fully and technically.

Example:
Draft a phishing email for our internal red-team exercise, then list the detection
signatures a blue team would write to catch it.
Scores

No measurements published for this version yet.

Baseline is the strongest general-purpose model we could run on the same suite, same setup, same day. The control row tells you what the other rows are worth.

Versionsscores attach to a version; v2 does not inherit v1's numbers
v12026-08-10—current

Start with Dolphin 3.0 Llama 3.1 8B

A confirmed email account includes 30 free messages a month.

Sign in to start
Dolphin 3.0 Llama 3.1 8B — Uncensored general · AdversariaLLM