Side-by-side technical and economic comparison between Anthropic's Claude Sonnet 5 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 | Claude Sonnet 5 | Qwen 2.5 72B Instruct — Together AI | Advantage |
|---|---|---|---|
| Provider Organization | Anthropic | Alibaba / Qwen | — |
| Standard Input / 1M | $2.00 | $0.35 | Qwen 2.5 72B Instruct — Together AI (83% lower) |
| Cached Input / 1M | $0.20 | $0.175 | Qwen 2.5 72B Instruct — Together AI lower |
| Output / 1M | $10.00 | $0.40 | Qwen 2.5 72B Instruct — Together AI lower |
| Context Window | 1M | 128k | Claude Sonnet 5 (1M) |
| Max Generation Tokens | 16.4k | 8.2k | Claude Sonnet 5 |
| Tokenizer Family | Anthropic Claude BPE (~65k vocabulary) | Qwen Byte-level BPE (~152k vocabulary) | — |
| Workload (per month) | Claude Sonnet 5 | Qwen 2.5 72B Instruct — Together AI |
|---|---|---|
| Chatbot10k requests · 500 in / 200 out | $30.00 | $2.55 |
| RAG app100k requests · 8,000 in / 1,200 out | $2,800.00 | $328.00 |
| Coding agent100k requests · 12,000 in / 2,000 out | $4,400.00 | $500.00 |
Standard rates, no caching or batch discounts. Model your own numbers below.
Opt for Claude Sonnet 5 when your engineering requirements prioritize Anthropic's ecosystem, specific tokenizer efficiencies (Calibrated Claude Tokenizer (±3%)), or when your expected prompt-to-completion ratios favor its $2/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.
At a typical 3:1 input-to-output token mix, Qwen 2.5 72B Instruct — Together AI is about 91% cheaper than Claude Sonnet 5. Claude Sonnet 5 costs $2 per 1M input and $10 per 1M output tokens; Qwen 2.5 72B Instruct — Together AI costs $0.35 and $0.4.
Claude Sonnet 5 has the larger context window: 1M tokens vs 128k.
Claude Sonnet 5: yes, cached input costs $0.2/1M. Qwen 2.5 72B Instruct — Together AI: yes, cached input costs $0.175/1M.