Side-by-side comparison of pricing, 12 benchmarks, and generation speed.
| Metric | Llama 3.2 Instruct 90B (Vision) | Qwen3.6 35B A3B (Reasoning) |
|---|---|---|
| Input ($/M tokens) | $1.38 | $0.248 |
| Output ($/M tokens) | $1.38 | $1.485 |
Data from Artificial Analysis API — 12 benchmarks
Qwen3.6 35B A3B (Reasoning) is cheaper overall. Its blended price (3:1 input/output ratio) is $0.56/M tokens vs $1.38/M for Llama 3.2 Instruct 90B (Vision).
Qwen3.6 35B A3B (Reasoning) wins 7 out of 12 benchmarks compared to 3 for Llama 3.2 Instruct 90B (Vision). See the detailed benchmark chart above for per-category results.
Qwen3.6 35B A3B (Reasoning) generates tokens faster at 188 tok/s vs 43 tok/s. Llama 3.2 Instruct 90B (Vision) also has lower time-to-first-token (0.59s vs 1.46s).
Choose based on your priorities: Qwen3.6 35B A3B (Reasoning) for lower cost, Qwen3.6 35B A3B (Reasoning) for stronger benchmark performance, and Qwen3.6 35B A3B (Reasoning) for faster generation. For latency-sensitive apps, check the TTFT comparison above.