Llama 3.3 70B SpecDec (Groq) token cost

Speculative-decoding oriented Llama 3.3 70B path on Groq when the host exposes it for lower latency. Compare against standard versatile 70B pricing and quality.

Privacy first counting

TokenCALC counts tokens in your browser when an Exact encoding is available. Prompt text does not need to leave your device for local counting.

Pricing snapshot

USD per 1 million tokens. Verify against the provider source before you lock budget.

RateUSD
Input / 1M$0.590
Output / 1M$0.990

Model details

DetailValue
ProviderGroq
Context window8.2K
TokenizerApprox
Last verified2026-08-02
API model idllama-3.3-70b-specdec