**Root Cause:**
The requested context window (**262,144 tokens**) is too large for the available GPU memory (**1,024 MiB**). The KV cache (the memory used to store the "memory" of the conversation) scales with context length; a 256k context window requires several gigabytes of VRAM even with aggressive quantization, far exceeding your 1GB of free space.

**Next Two Actions:**
1. **Reduce Context Size:** Lower the context window parameter (e.g., use `-c 8192` or `-c 32768`) to a size that fits within your available VRAM.
2. **Use a Smaller/More Quantized Model:** Switch to a smaller parameter model (e.g., 1B–3B) or a more aggressive quantization (e.g., Q4_K_S or Q2_K) to reduce the base model's memory footprint, leaving more room for the KV cache.