The closest alternative to Llama 3 70B Instruct is Gemma 3 12B — comparable intelligence (8.8 vs 8.9); for a cheaper option, MiniMax M3 runs $0.300/M vs $0.510/M; for more speed, gpt-oss-120b (free) hits 554 t/s; and the best open-weight substitute is MiniMax M3.
Gemma 3 12B is the closest alternative to Llama 3 70B Instruct, with an Intelligence Index of 8.8 (Llama 3 70B Instruct scores 8.9) at $0.050 per million input tokens.
Yes — MiniMax M3 costs $0.300 per million input tokens versus $0.510 for Llama 3 70B Instruct, while staying within ~8 points of its Intelligence Index.
MiniMax M3 is the strongest open-weight alternative to Llama 3 70B Instruct — downloadable from Hugging Face and self-hostable, scoring 54.7 on intelligence.