Side-by-side technical and economic comparison between Mistral AI's Mistral Large 3 and Alibaba / Qwen's Qwen 2.5 72B Instruct — Together AI.
• Qwen 2.5 72B Instruct — Together AI: Classified as a Legacy Reference model (Deprecated on Together AI (Reference)).
| Metric | Mistral Large 3 | Qwen 2.5 72B Instruct — Together AI | Advantage |
|---|---|---|---|
| Provider Organization | Mistral AI | Alibaba / Qwen | — |
| Standard Input / 1M | $0.50 | $0.35 | Qwen 2.5 72B Instruct — Together AI (30% lower) |
| Cached Input / 1M | $0.05 | $0.175 | Mistral Large 3 lower |
| Output / 1M | $1.50 | $0.40 | Qwen 2.5 72B Instruct — Together AI lower |
| Context Window | 256k | 128k | Mistral Large 3 (256k) |
| Max Generation Tokens | 16.4k | 8.2k | Mistral Large 3 |
| Tokenizer Family | Mistral Tekken Tokenizer (131k vocabulary) | Qwen Byte-level BPE (~152k vocabulary) | — |
Opt for Mistral Large 3 when your engineering requirements prioritize Mistral AI's ecosystem, specific tokenizer efficiencies (Calibrated Mistral Tokenizer (±3%)), or when your expected prompt-to-completion ratios favor its $0.5/M input rate.
Opt for Qwen 2.5 72B Instruct — Together AI when looking for Alibaba / Qwen's tooling integration, specific context window depth (128k tokens), or when output generation volume favors its $0.4/M rate.