| Small Models | |||||
| Model Name | Version | Context | Speed* | Prompt Format | Recommendation |
|---|---|---|---|---|---|
| Meta-Llama-3.1-8B-Instruct[NEW] | 3.1 | 131,072 | SR: 50t/s PR: 400t/s | Llama 3 Instruct | Meta's Newest Llama model. Great for general use. |
| Meta-Llama-3-8B-Instruct | 3.0 Abliterated | 8,192 | SR: 50t/s PR: 400t/s | Llama 3 Instruct | Great for general use, verbose, abliterated for less refusal |
| Awanllm-Llama-3-8B-Dolfin | 1.0 | 8,192 | SR: 50t/s PR: 400t/s | Llama 3 Instruct | Exact instructions following, less refusals, no warnings |
| Awanllm-Llama-3-8B-Cumulus | 1.0 | 8,192 | SR: 50t/s PR: 400t/s | Llama 3 Instruct | Great for storywriting or RP, zero refusal, follows characters |
| Large Models | |||||
| Model Name | Version | Context | Speed* | Prompt Format | Recommendation |
| Meta-Llama-3.1-70B-Instruct[NEW] | 3.1 | 131,072 | SR: 20t/s PR: 100t/s | Llama 3 Instruct | Great for general use, SOTA dense LLM model |
| Meta-Llama-3-70B-Instruct | 1.0 | 8,192 | SR: 20t/s PR: 100t/s | Llama 3 Instruct | Great for general use, SOTA dense LLM model |