Side-by-side comparison of pricing, 12 benchmarks, and generation speed.
| Metric | Qwen3 4B (Non-reasoning) | DeepSeek V3.2 Exp (Reasoning) |
|---|---|---|
| Input ($/M tokens) | $0.11 | $0.28 |
| Output ($/M tokens) | $0.42 | $0.42 |
Data from Artificial Analysis API — 12 benchmarks
Qwen3 4B (Non-reasoning) is cheaper overall. Its blended price (3:1 input/output ratio) is $0.19/M tokens vs $0.32/M for DeepSeek V3.2 Exp (Reasoning).
DeepSeek V3.2 Exp (Reasoning) wins 11 out of 12 benchmarks compared to 1 for Qwen3 4B (Non-reasoning). See the detailed benchmark chart above for per-category results.
Qwen3 4B (Non-reasoning) generates tokens faster at 105 tok/s vs 32 tok/s. Qwen3 4B (Non-reasoning) also has lower time-to-first-token (0.93s vs 1.48s).
Choose based on your priorities: Qwen3 4B (Non-reasoning) for lower cost, DeepSeek V3.2 Exp (Reasoning) for stronger benchmark performance, and Qwen3 4B (Non-reasoning) for faster generation. For latency-sensitive apps, check the TTFT comparison above.