DeepSeek R1 0528 Qwen3 8B vs Gemma 4 26B A4B IT
Price and specs side by side · USD per 1M tokens · verified 2026-10-11
// side by side
| metric | DeepSeek R1 0528 Qwen3 8B | Gemma 4 26B A4B IT |
|---|
| input / 1M | $0.06 | $0.07 |
| output / 1M | $0.09 | $0.34 |
| blended (3:1) | $0.0675 | $0.1375 |
| cache read / 1M | — | — |
| context | 128K | 262.1K |
| max output | 32K | 32.8K |
| input types | text | text · image |
| reasoning | yes | yes |
| tool calling | no | yes |
| structured output | no | yes |
| open weights | yes | yes |
| released | 2025-05-29 | 2026-04-02 |
| knowledge cutoff | n/a | n/a |
| lab | DeepSeek | Google |
// monthly cost scenarios
| workload | tokens in / out | DeepSeek R1 0528 Qwen3 8B | Gemma 4 26B A4B IT |
|---|
| Chatbot | 10M / 2M | $0.78 | $1.38 |
| RAG app | 50M / 5M | $3.45 | $5.20 |
| Coding agent | 20M / 10M | $2.10 | $4.80 |
// verdict
- DeepSeek R1 0528 Qwen3 8B is 50.9% cheaper than Gemma 4 26B A4B IT on a blended 3:1 input/output basis.
- Gemma 4 26B A4B IT has the larger context window (262.1K vs 128K tokens).
- Only Gemma 4 26B A4B IT supports tool calling.
- Only Gemma 4 26B A4B IT supports structured output.
- Only Gemma 4 26B A4B IT accepts image input.