GPT-5.4 vs Qwen3.5-397B-A17B

On shared benchmark coverage, MMLU PRO has GPT-5.4 at 87.5 and Qwen3.5-397B-A17B at 87.8, with Qwen3.5-397B-A17B ahead by 0.3 points; Google-Proof Q&A has GPT-5.4 at 92 and Qwen3.5-397B-A17B at 89.3, with GPT-5.4 ahead by 2.7 points; Massive Multi-discipline Multimodal Understanding has GPT-5.4 at 82.1 and Qwen3.5-397B-A17B at 85, with Qwen3.5-397B-A17B ahead by 2.9 points. The largest visible gap is 2.9 points on Massive Multi-discipline Multimodal Understanding, which matters most when that benchmark mirrors your workload. Treat isolated benchmark wins as directional, because provider routing, prompt style, and tool access can move real application results.

The capability footprint differs most on reasoning mode: GPT-5.4, function calling: GPT-5.4, tool use: GPT-5.4, and code execution: GPT-5.4. Both models share multimodal input and structured outputs, so the practical split is not just feature count. Use those differences to decide whether the page is about raw model quality, agentic coding support, multimodal ingestion, or predictable structured API behavior.

For cost, GPT-5.4 lists $2.5/1M input and $15/1M output tokens, while Qwen3.5-397B-A17B lists $0.39/1M input and $2.34/1M output tokens on the cheapest tracked provider. A 70/30 input-output blend puts Qwen3.5-397B-A17B lower by about $5.28 per million blended tokens. Availability is 2 providers versus 1, so concentration risk also matters.

Choose GPT-5.4 when coding workflow support and broader provider choice are central to the workload. Choose Qwen3.5-397B-A17B when provider fit and lower input-token cost are more important. For production, rerun your own prompts through the exact provider, region, and tool stack you plan to ship.