Strip <think>...</think> blocks from reasoning model output (e.g. Qwen3-VL-Thinking) and increase max_tokens from 4096 to 16384 to accommodate thinking token overhead. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Strip <think>...</think> blocks from reasoning model output (e.g. Qwen3-VL-Thinking) and increase max_tokens from 4096 to 16384 to accommodate thinking token overhead. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>