Side-by-side technical and economic comparison between xAI's Grok 2 and Meta / Together's Llama 3.1 405B — Together AI.
• Grok 2: Classified as a Legacy Reference model (Generally Available).
• Llama 3.1 405B — Together AI: Classified as a Legacy Reference model (Deprecated on Together AI (Reference)).
| Metric | Grok 2 | Llama 3.1 405B — Together AI | Advantage |
|---|---|---|---|
| Provider Organization | xAI | Meta / Together | — |
| Standard Input / 1M | $2.00 | $3.50 | Grok 2 (43% lower) |
| Cached Input / 1M | None | None | Llama 3.1 405B — Together AI lower |
| Output / 1M | $10.00 | $3.50 | Llama 3.1 405B — Together AI lower |
| Context Window | 128k | 128k | Equal |
| Max Generation Tokens | 8.2k | 4.1k | Grok 2 |
| Tokenizer Family | xAI Byte-level BPE (~131k vocabulary) | Meta Llama 3 Tiktoken BPE (~128k vocabulary) | — |
Opt for Grok 2 when your engineering requirements prioritize xAI's ecosystem, specific tokenizer efficiencies (Calibrated xAI Tokenizer (±3%)), or when your expected prompt-to-completion ratios favor its $2/M input rate.
Opt for Llama 3.1 405B — Together AI when looking for Meta / Together's tooling integration, specific context window depth (128k tokens), or when output generation volume favors its $3.5/M rate.