Dataset Viewer
The dataset viewer is not available for this split.
Cannot load the dataset split (in streaming mode) to extract the first rows.
Error code: StreamingRowsError
Exception: CastError
Message: Couldn't cast
ts: double
task_id: string
task_name: string
reasoning_level: string
model_set: string
reasoning_overrides: struct<mtplx: string>
child 0, mtplx: string
harnesses: list<item: string>
child 0, item: string
frameworks_run: list<item: string>
child 0, item: string
settings: struct<temperature: double, max_tokens: int64, cell_timeout_s: int64, top_p: double>
child 0, temperature: double
child 1, max_tokens: int64
child 2, cell_timeout_s: int64
child 3, top_p: double
frameworks: struct<omlx: struct<port: int64, model: string>, mtplx: struct<port: int64, model: string>, mlxvlm: (... 123 chars omitted)
child 0, omlx: struct<port: int64, model: string>
child 0, port: int64
child 1, model: string
child 1, mtplx: struct<port: int64, model: string>
child 0, port: int64
child 1, model: string
child 2, mlxvlm: struct<port: int64, model: string>
child 0, port: int64
child 1, model: string
child 3, mlxserve: struct<port: int64, model: string>
child 0, port: int64
child 1, model: string
child 4, vllm: struct<port: int64, model: string>
child 0, port: int64
child 1, model: string
results: list<item: struct<framework: string, model: string, harness: string, task: string, status: string, l (... 583 chars omitted)
child 0, item: struct<framework: string, model: string, harness: string, task: string, status: string, latency: dou (... 571 chars omitted)
child 0, framework: string
child 1, model: stri
...
d: bool
child 10, tokens_estimated: bool
child 11, pp: double
child 12, tgs: double
child 13, ttft: double
child 14, prompt_tokens: int64
child 15, wall: double
child 16, iterations: int64
child 17, server_pp: double
child 18, server_tgs: double
child 19, server_tgs_flagged: bool
child 20, server_prompt_tokens: int64
child 21, server_completion_tokens: int64
child 22, qa_score: int64
child 23, qa_notes: string
child 24, ctx_fill_pct: double
child 25, est_gbps: double
child 26, continuations: int64
child 27, calls: int64
child 28, tokens_source: string
child 29, hart_status: string
child 30, server_requests: int64
child 31, server_ttft_avg: double
child 32, server_pp_method: string
child 33, server_tgs_method: string
server_requests: int64
server_pp_method: string
tokens_source: string
qa_score: int64
calls: int64
pp: double
continuations: int64
qa_notes: string
server_tgs_method: string
ttft: double
hart_status: string
server_tgs: double
prompt_tokens: int64
tokens_estimated: bool
output_url: string
tps: double
task: string
tokens: int64
server_ttft_avg: double
framework: string
model: string
server_pp: double
truncated: bool
tgs: double
server_tgs_flagged: bool
est_gbps: double
iterations: int64
wall: double
ctx_fill_pct: double
latency: double
harness: string
status: string
server_prompt_tokens: int64
server_completion_tokens: int64
to
{'framework': Value('string'), 'model': Value('string'), 'harness': Value('string'), 'task': Value('string'), 'status': Value('string'), 'latency': Value('float64'), 'tokens': Value('int64'), 'tps': Value('float64'), 'output_url': Value('string'), 'truncated': Value('bool'), 'tokens_estimated': Value('bool'), 'pp': Value('float64'), 'tgs': Value('float64'), 'ttft': Value('float64'), 'prompt_tokens': Value('int64'), 'wall': Value('float64'), 'iterations': Value('int64'), 'server_pp': Value('float64'), 'server_tgs': Value('float64'), 'server_tgs_flagged': Value('bool'), 'server_prompt_tokens': Value('int64'), 'server_completion_tokens': Value('int64'), 'qa_score': Value('int64'), 'qa_notes': Value('string'), 'ctx_fill_pct': Value('float64'), 'est_gbps': Value('float64'), 'continuations': Value('int64'), 'calls': Value('int64'), 'tokens_source': Value('string'), 'hart_status': Value('string'), 'server_requests': Value('int64'), 'server_ttft_avg': Value('float64'), 'server_pp_method': Value('string'), 'server_tgs_method': Value('string')}
because column names don't match
Traceback: Traceback (most recent call last):
File "/src/services/worker/src/worker/utils.py", line 147, in get_rows_or_raise
return get_rows(
dataset=dataset,
...<4 lines>...
column_names=column_names,
)
File "/src/libs/libcommon/src/libcommon/utils.py", line 272, in decorator
return func(*args, **kwargs)
File "/src/services/worker/src/worker/utils.py", line 127, in get_rows
rows_plus_one = list(itertools.islice(safe_iter(ds, dataset=dataset), rows_max_number + 1))
File "/src/services/worker/src/worker/utils.py", line 483, in safe_iter
yield from ds.decode(False) if ds.features else ds
File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 2951, in __iter__
for key, example in ex_iterable:
^^^^^^^^^^^
File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 2461, in __iter__
for key, pa_table in self._iter_arrow():
~~~~~~~~~~~~~~~~^^
File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 2486, in _iter_arrow
for key, pa_table in self.ex_iterable._iter_arrow():
~~~~~~~~~~~~~~~~~~~~~~~~~~~~^^
File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 547, in _iter_arrow
for key, pa_table in iterator:
^^^^^^^^
File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 430, in _iter_arrow
for key, pa_table in self.generate_tables_fn(**gen_kwags):
~~~~~~~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^
File "/usr/local/lib/python3.14/site-packages/datasets/packaged_modules/json/json.py", line 343, in _generate_tables
self._cast_table(pa_table, json_field_paths=json_field_paths),
~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/usr/local/lib/python3.14/site-packages/datasets/packaged_modules/json/json.py", line 132, in _cast_table
pa_table = table_cast(pa_table, self.info.features.arrow_schema)
File "/usr/local/lib/python3.14/site-packages/datasets/table.py", line 2378, in table_cast
return cast_table_to_schema(table, schema)
File "/usr/local/lib/python3.14/site-packages/datasets/table.py", line 2306, in cast_table_to_schema
raise CastError(
...<3 lines>...
)
datasets.table.CastError: Couldn't cast
ts: double
task_id: string
task_name: string
reasoning_level: string
model_set: string
reasoning_overrides: struct<mtplx: string>
child 0, mtplx: string
harnesses: list<item: string>
child 0, item: string
frameworks_run: list<item: string>
child 0, item: string
settings: struct<temperature: double, max_tokens: int64, cell_timeout_s: int64, top_p: double>
child 0, temperature: double
child 1, max_tokens: int64
child 2, cell_timeout_s: int64
child 3, top_p: double
frameworks: struct<omlx: struct<port: int64, model: string>, mtplx: struct<port: int64, model: string>, mlxvlm: (... 123 chars omitted)
child 0, omlx: struct<port: int64, model: string>
child 0, port: int64
child 1, model: string
child 1, mtplx: struct<port: int64, model: string>
child 0, port: int64
child 1, model: string
child 2, mlxvlm: struct<port: int64, model: string>
child 0, port: int64
child 1, model: string
child 3, mlxserve: struct<port: int64, model: string>
child 0, port: int64
child 1, model: string
child 4, vllm: struct<port: int64, model: string>
child 0, port: int64
child 1, model: string
results: list<item: struct<framework: string, model: string, harness: string, task: string, status: string, l (... 583 chars omitted)
child 0, item: struct<framework: string, model: string, harness: string, task: string, status: string, latency: dou (... 571 chars omitted)
child 0, framework: string
child 1, model: stri
...
d: bool
child 10, tokens_estimated: bool
child 11, pp: double
child 12, tgs: double
child 13, ttft: double
child 14, prompt_tokens: int64
child 15, wall: double
child 16, iterations: int64
child 17, server_pp: double
child 18, server_tgs: double
child 19, server_tgs_flagged: bool
child 20, server_prompt_tokens: int64
child 21, server_completion_tokens: int64
child 22, qa_score: int64
child 23, qa_notes: string
child 24, ctx_fill_pct: double
child 25, est_gbps: double
child 26, continuations: int64
child 27, calls: int64
child 28, tokens_source: string
child 29, hart_status: string
child 30, server_requests: int64
child 31, server_ttft_avg: double
child 32, server_pp_method: string
child 33, server_tgs_method: string
server_requests: int64
server_pp_method: string
tokens_source: string
qa_score: int64
calls: int64
pp: double
continuations: int64
qa_notes: string
server_tgs_method: string
ttft: double
hart_status: string
server_tgs: double
prompt_tokens: int64
tokens_estimated: bool
output_url: string
tps: double
task: string
tokens: int64
server_ttft_avg: double
framework: string
model: string
server_pp: double
truncated: bool
tgs: double
server_tgs_flagged: bool
est_gbps: double
iterations: int64
wall: double
ctx_fill_pct: double
latency: double
harness: string
status: string
server_prompt_tokens: int64
server_completion_tokens: int64
to
{'framework': Value('string'), 'model': Value('string'), 'harness': Value('string'), 'task': Value('string'), 'status': Value('string'), 'latency': Value('float64'), 'tokens': Value('int64'), 'tps': Value('float64'), 'output_url': Value('string'), 'truncated': Value('bool'), 'tokens_estimated': Value('bool'), 'pp': Value('float64'), 'tgs': Value('float64'), 'ttft': Value('float64'), 'prompt_tokens': Value('int64'), 'wall': Value('float64'), 'iterations': Value('int64'), 'server_pp': Value('float64'), 'server_tgs': Value('float64'), 'server_tgs_flagged': Value('bool'), 'server_prompt_tokens': Value('int64'), 'server_completion_tokens': Value('int64'), 'qa_score': Value('int64'), 'qa_notes': Value('string'), 'ctx_fill_pct': Value('float64'), 'est_gbps': Value('float64'), 'continuations': Value('int64'), 'calls': Value('int64'), 'tokens_source': Value('string'), 'hart_status': Value('string'), 'server_requests': Value('int64'), 'server_ttft_avg': Value('float64'), 'server_pp_method': Value('string'), 'server_tgs_method': Value('string')}
because column names don't matchNeed help to make the dataset viewer work? Make sure to review how to configure the dataset viewer, and open a discussion for direct support.
YAML Metadata Warning:The task_categories "benchmarking" is not in the official list: text-classification, token-classification, table-question-answering, question-answering, zero-shot-classification, translation, summarization, feature-extraction, text-generation, fill-mask, sentence-similarity, text-to-speech, text-to-audio, automatic-speech-recognition, audio-to-audio, audio-classification, audio-text-to-text, voice-activity-detection, depth-estimation, image-classification, object-detection, image-segmentation, text-to-image, image-to-text, image-to-image, image-to-video, unconditional-image-generation, video-classification, reinforcement-learning, robotics, tabular-classification, tabular-regression, tabular-to-text, table-to-text, multiple-choice, text-ranking, text-retrieval, time-series-forecasting, text-to-video, image-text-to-text, image-text-to-image, image-text-to-video, visual-question-answering, document-question-answering, zero-shot-image-classification, graph-ml, mask-generation, zero-shot-object-detection, text-to-3d, image-to-3d, image-feature-extraction, video-text-to-text, keypoint-detection, visual-document-retrieval, any-to-any, video-to-video, other
benchtest results — Apple Silicon framework × harness benchmark
Raw benchmark data from benchtest: one Mac (Apple M5 Max, 128 GB), local LLM serving frameworks (OMLX / MTPLX / MLX-VLM / MLX-Serve) crossed with six agent harnesses (raw API, raw+, pi, opencode, aider, hart) on three build tasks (tetris, code review, agent console).
Datasets
| Folder | Model set | Cells | Done |
|---|---|---|---|
qwen38-27b/ |
Qwen3.8-27B, one build per framework | 72 | 72 |
flash-next/ |
Qwen3.8 Flash-Next (MTP) builds | 54 | 53 |
Each folder holds one JSON per task run (per-cell records: framework, harness, model actually served, status, latency, tokens, TPS, server-measured PP/TGS, iterations, QA reliability score) plus per-cell JSONL crash-safety logs and a manifest README.
Headline finding
Flash-Next (MTP) is ~1.7× faster on MTPLX (84.9 tok/s) and OMLX (87.6) but a wash on MLX-VLM (39.0) — the framework decides whether the model architecture pays off. MLX-Serve cannot load the flash-next build at all on 128 GB (memory preflight needs ~112.5 GB).
Metric dictionary
tps— overall tokens/wall-second (client); for agent harnesses this includes tool-execution timeserver_tgs/server_pp— server-measured decode / prefill tok/s (the honest cross-harness speed metric)iterations— model calls per cellqa_score— deterministic artifact reliability checks (0–100); quality judgment stays humantokens_source: server— token count from the framework's own counters
Generated by benchtest v1.0 · 2026-10.
- Downloads last month
- 58