Qwen3.6-35B-A3B tool calling benchmark: ByteShape vs. Unsloth GGUFs, KV cache quants & long context performance

In three linesQualitative tool calling benchmark on Qwen3.6-35B-A3B comparing ByteShape vs. Unsloth quantizations, with KV cache tests (q8_0, q4_0) and long context. Results: no clear ByteShape/Unsloth winner; q8_0 is free lunch, q4_0 degrades performance; long context significantly reduces tool calling across all scenarios.

Summary generated by Claude — human-verified