feat: add tool-calling run for mlx-community/Qwen3-Next-80B-A3B-Thinking-4bit

feat: add tool-calling run for mlx-community/Qwen3-Next-80B-A3B-Thinking-4bit

feat: add tool-calling run for mlx-community/Qwen3-Next-80B-A3B-Thinking-4bit

feat: add tool-calling run for mlx-community/Qwen3-Next-80B-A3B-Thinking-4bit

feat: add coding run for mlx-community/Qwen3-Coder-30B-A3B-Instruct-4bit

feat: add tool-calling run for lmstudio-community/Hermes-4-70B-MLX-4bit

feat: add tool-calling run for OpenYourMind/Qwopus3.5-122B-A10B-Kimi-K2.6-destill-healed-abliterated-MTPLX-4bit

feat: add tool-calling run for mlx-community/gpt-oss-120b-MXFP4-Q8

feat: add tool-calling run for mlx-community/Qwen3-Coder-30B-A3B-Instruct-4bit

feat: add tool-calling run for mlx-community/Qwen3-Next-80B-A3B-Thinking-4bit

feat: add tool-calling run for lmstudio-community/Qwen3.6-35B-A3B-MLX-8bit

feat: add tool-calling run for mlx-community/Qwen3.6-35B-A3B-8bit

feat: add tool-calling run for mlx-community/GLM-4.7-Flash-4bit

feat: add tool-calling run for mlx-community/Qwen3.6-35B-A3B-4bit

feat: add tool-calling run for mlx-community/Qwen3.6-35B-A3B-OptiQ-4bit

feat: add reasoning run for openrouter/openrouter/auto

feat: add reasoning run for gemini/gemini-2.5-flash

feat: add math-hard run for mlx-community/Qwen3-Coder-30B-A3B-Instruct-4bit

chore(data): prune superseded dead-quant shards (Qwen3.5-*, Qwen3.6 mxfp4/mxfp8)

feat: add throughput run for mlx-community/Qwen3-Coder-30B-A3B-Instruct-4bit

feat: add throughput run for mlx-community/Qwen3-Coder-30B-A3B-Instruct-4bit

feat: add throughput run for mlx-community/Qwen3-Coder-30B-A3B-Instruct-4bit

feat: add throughput run for mlx-community/gpt-oss-120b-MXFP4-Q8

feat: add throughput run for mlx-community/gpt-oss-120b-MXFP4-Q8

feat: add throughput run for mlx-community/gpt-oss-120b-MXFP4-Q8

feat: add coding run for mlx-community/Qwen3.6-35B-A3B-mxfp4

feat(backfill): populate duration_seconds and tokens-per-second on historical runs

feat: publish 12 benchmark runs (local sweep 2026-04-19)

feat: rescue 12 older orphan benchmark files (12 rows from 2026-03-23 and 2026-03-30 runs)

feat: rescue 26 orphan F8 sweep files (91 rows, 13 models x 2 suites)

test(scaffold-poc): remove after successful gating verification

test(scaffold-poc): verify HfApi append pattern with unique filename

docs: remove dangling repo reference, generalize voice

feat: initial import of 140 benchmark rows from 56 historical runs

docs: add reference envelope schema (v1)

chore: add .gitattributes for Parquet LFS

docs: add dataset card with schema, quickstart, and configs block

initial commit