Side-by-side comparison of pricing, 12 benchmarks, and generation speed.
| Metric | Qwen3.5 397B A17B (Non-reasoning) | DeepSeek V4 Pro 0813 (Reasoning, Max Effort) |
|---|---|---|
| Input ($/M tokens) | $0.6 | $1.32 |
| Output ($/M tokens) | $3.6 | $3.96 |
Verdict. DeepSeek V4 Pro 0813 (Reasoning, Max Effort) takes the aggregate benchmark matchup 2–0 across 2 categories. Real workloads usually care about a handful of specific tasks — see the per-benchmark table above.
Pricing. Both models sit in the budget bracket for output-token pricing. At 0.9× the per-million-token cost, Qwen3.5 397B A17B (Non-reasoning) is meaningfully cheaper if your traffic is output-heavy (long completions, document generation, agent loops). Qwen3.5 397B A17B (Non-reasoning) makes more sense when output volume is low and absolute reasoning quality justifies the premium.
Strengths. Qwen3.5 397B A17B (Non-reasoning) is strongest on Intelligence Index (32.7). DeepSeek V4 Pro 0813 (Reasoning, Max Effort) leads on Coding Index (68.8), Intelligence Index (53.2).
Speed. Throughput is comparable — 83 tok/s vs 74 tok/s — so generation speed shouldn't drive your choice here. Look at the per-benchmark wins instead.
Provider. Alibaba and DeepSeek sell to overlapping but distinct developer audiences: Alibaba tends to ship frontier reasoning models with premium positioning, while DeepSeek often prices more aggressively. Your existing vendor relationships, billing, and SLA preferences may matter as much as the raw numbers above.
Workload cost. Workload scenarios (per million requests at 30M input + 15M output tokens): Qwen3.5 397B A17B (Non-reasoning) costs $72.00 ($864/year); DeepSeek V4 Pro 0813 (Reasoning, Max Effort) costs $99.00 ($1188/year). At a smaller 5M-input/2M-output scale (single-developer tool or prototype): Qwen3.5 397B A17B (Non-reasoning) ≈ $10.20/run, DeepSeek V4 Pro 0813 (Reasoning, Max Effort) ≈ $14.52/run. At agent/realtime scale (200M input / 100M output per million requests): Qwen3.5 397B A17B (Non-reasoning) ≈ $480/run, DeepSeek V4 Pro 0813 (Reasoning, Max Effort) ≈ $660/run. Qwen3.5 397B A17B (Non-reasoning) becomes more attractive at higher volume — the absolute per-token pricing difference compounds when you ship at scale.
Recommendation. Both models have legitimate use cases — the right answer depends on whether you are optimizing for benchmark ceiling, latency, or unit cost. Start with the cheaper / faster model, evaluate against your specific task, and only switch if the upgrade shows a meaningful lift.
Data from Artificial Analysis API — 12 benchmarks
Qwen3.5 397B A17B (Non-reasoning) is cheaper overall. Its blended price (3:1 input/output ratio) is $1.35/M tokens vs $1.98/M for DeepSeek V4 Pro 0813 (Reasoning, Max Effort).
DeepSeek V4 Pro 0813 (Reasoning, Max Effort) wins 2 out of 12 benchmarks compared to 0 for Qwen3.5 397B A17B (Non-reasoning). See the detailed benchmark chart above for per-category results.
Qwen3.5 397B A17B (Non-reasoning) generates tokens faster at 83 tok/s vs 74 tok/s. However, DeepSeek V4 Pro 0813 (Reasoning, Max Effort) has lower time-to-first-token (1.70s vs 2.04s).
Choose based on your priorities: Qwen3.5 397B A17B (Non-reasoning) for lower cost, DeepSeek V4 Pro 0813 (Reasoning, Max Effort) for stronger benchmark performance, and Qwen3.5 397B A17B (Non-reasoning) for faster generation. For latency-sensitive apps, check the TTFT comparison above.