DeepSeek R1 0528 Qwen3 8B vs Voxtral Small 24B 2507
Price and specs side by side · USD per 1M tokens · verified 2026-10-11
// side by side
| metric | DeepSeek R1 0528 Qwen3 8B | Voxtral Small 24B 2507 |
|---|
| input / 1M | $0.06 | $0.10 |
| output / 1M | $0.09 | $0.30 |
| blended (3:1) | $0.0675 | $0.15 |
| cache read / 1M | — | — |
| context | 128K | 32.8K |
| max output | 32K | 8.2K |
| input types | text | text · audio |
| reasoning | yes | no |
| tool calling | no | yes |
| structured output | no | yes |
| open weights | yes | yes |
| released | 2025-05-29 | 2025-07-15 |
| knowledge cutoff | n/a | n/a |
| lab | DeepSeek | Mistral |
// monthly cost scenarios
| workload | tokens in / out | DeepSeek R1 0528 Qwen3 8B | Voxtral Small 24B 2507 |
|---|
| Chatbot | 10M / 2M | $0.78 | $1.60 |
| RAG app | 50M / 5M | $3.45 | $6.50 |
| Coding agent | 20M / 10M | $2.10 | $5.00 |
// verdict
- DeepSeek R1 0528 Qwen3 8B is 55% cheaper than Voxtral Small 24B 2507 on a blended 3:1 input/output basis.
- DeepSeek R1 0528 Qwen3 8B has the larger context window (128K vs 32.8K tokens).
- Only DeepSeek R1 0528 Qwen3 8B supports built-in reasoning.
- Only Voxtral Small 24B 2507 supports tool calling.
- Only Voxtral Small 24B 2507 supports structured output.