Llama 3.3 70B Instruct (Fast)

Meta · reasoning · 24K context · 3× credits

Meta's flagship 70B instruct model. Strong general reasoning, coding and tool calling.

Model ID: mixerlead-ai/meta/llama-3.3-70b-instruct-fp8-fast

Best for

Meta's large 70B model. Very good at following detailed instructions and using tools.

AI agents that must call the right function with the right inputs

Tasks with many rules or a strict output format

Clean JSON output with response_format={"type": "json_object"}

Context is 24K tokens, so it is not the best choice for very long documents (use Llama 4 Scout).

It costs 3× credits. For cheaper reasoning, try GPT-OSS 120B (0.75×).