Provenance-stamped · OSS origin/main @ ae1d381

The benchmark leaderboard explorer

The full benchmark corpus — every model, scenario cell, and config the launch leaderboard measured — laid out as one flat, faceatable dataset. Narrow it by model, role, cell, config, lane, variant, or discretion arm; pivot model × cell or config × cell; sort by α, α-per-dollar, arm rate, or latency; and drill any row to its numbers and its docs/results/… provenance. Every figure is transcribed from a committed OSS result file, never a live query.

Honesty is structural, not cosmetic: latency is measured but marked UNPRICED(it is not folded into α, so a slow-frontier “win” reads as an artifact of an unpriced axis, not a finding); an abstention or a range is never shown as a coerced $0; a lane is a neutral capability class, never a provider path; and the PM tier appears with a count of 0 because Morning-Book-v0 is unbuilt — we do not fabricate a PM row.

45 of 45 rows

Model

Role

Cell

Config

Lane

Variant

Discretion

Model	Role	Cell	Config	Lane	Variant	Discretion
best frontier model	Watcher	EQ-DAY-EARN	per-cell-discretion	frontier	best-frontier	discretion	+$1,627	n/a	n/a	n/a
gpt-5.4-mini	Watcher	EQ-DAY-EARN	earn-trap	small	—	flatten	+$1,627	n/a	n/a	n/a
Kimi K2.6	Watcher	EQ-DAY-EARN	open-reclaim	open	—	discretion	+$1,596 ± $0	7,528.3×	n/a	n/a
gpt-5	Watcher	EQ-DAY-EARN	earn-trap	frontier	—	flatten	+$1,570	n/a	n/a	n/a
best frontier model	Watcher	FOMC-DAY	per-cell-discretion	frontier	best-frontier	discretion	+$1,017	n/a	n/a	n/a
Qwen2.5-3B	Watcher	EQ-DAY-EARN	earn-trap	small	—	flatten	+$933	n/a	n/a	n/a
best open model	Watcher	FOMC-DAY	per-cell-discretion	open	best-open	discretion	+$773	n/a	n/a	n/a
best open model	Watcher	EQ-SCALP-ORB	per-cell-discretion	open	best-open	discretion	+$665	n/a	n/a	n/a
best frontier model	Watcher	EQ-SCALP-ORB	per-cell-discretion	frontier	best-frontier	discretion	+$665	n/a	n/a	n/a
best open model	Watcher	CTRL-QUIET-CHOP	per-cell-discretion	open	best-open	discretion	+$320	n/a	n/a	n/a
best frontier model	Watcher	CTRL-QUIET-CHOP	per-cell-discretion	frontier	best-frontier	discretion	+$320	n/a	n/a	n/a
GLM-5.2	Watcher	EQ-DAY-EARN	open-reclaim	open	—	discretion	+$231 ± $0	784.65×	n/a	n/a
qwen2.5-7b-local	Watcher	—	pair-discretion	local	tier-C	discretion	$205.75	n/a	n/a	n/a
claude-fable-5	Strategist	—	reasoning-level	frontier	medium	—	−$5.03	n/a	n/a	70.0s
claude-fable-5	Strategist	—	reasoning-level	frontier	high	—	−$30.34	n/a	n/a	103.0s
claude-fable-5	Strategist	—	reasoning-level	frontier	low	—	−$35.61	n/a	n/a	55.0s
Qwen2.5-7B	Watcher	EQ-DAY-EARN	cfg-ablation	small	cfg-off	discretion	−$97.80	n/a	n/a	n/a
claude-fable-5	Strategist	—	reasoning-level	frontier	max	—	−$142.59	n/a	n/a	99.0s
Qwen2.5-7B	Watcher	EQ-DAY-EARN	cfg-ablation	small	cfg-off	defined-risk	−$247.60	n/a	n/a	n/a
Qwen2.5-7B	Watcher	EQ-DAY-EARN	cfg-ablation	small	cfg-on	discretion	−$430.42	n/a	n/a	n/a
Qwen2.5-7B	Watcher	EQ-DAY-EARN	cfg-ablation	small	cfg-on	defined-risk	−$580.22	n/a	n/a	n/a
qwen3-max	Watcher	—	single-model	open	—	—	−$698.06	n/a	n/a	n/a
claude-sonnet-5	Watcher	—	single-model	frontier	—	—	−$707.66	n/a	n/a	n/a
gemini-3.1-pro	Watcher	—	single-model	frontier	—	—	−$713.00	n/a	n/a	n/a
claude-opus-4.8	Watcher	—	single-model	frontier	—	—	−$760.53	n/a	n/a	n/a
claude-haiku-4.5	Watcher	—	single-model	small	—	—	−$762.27	n/a	n/a	n/a
deepseek-v4-flash	Watcher	—	single-model	open	—	—	−$1,045.26	n/a	n/a	n/a
claude-opus-4.8 ≈ claude-haiku-4.5	Watcher	—	pair-discretion	frontier	tier-A	discretion	$846.25–$897.25	n/a	n/a	n/a
qwen3-max ≈ deepseek-v4-flash	Watcher	—	pair-discretion	open	tier-B	discretion	$377.80–$439.50	n/a	n/a	n/a
best open model	Watcher	EQ-DAY-EARN	per-cell-discretion	open	best-open	discretion	$0.00 (left money on the table)	n/a	n/a	n/a
Qwen2.5-7B	Watcher	EQ-DAY-EARN	earn-trap	small	—	—	$0.00 (no action)	n/a	n/a	n/a
Qwen2.5-0.5B	Watcher	EQ-DAY-EARN	size-sweep	small	—	—	−$29,748 ± $36,473	n/a	n/a	n/a
Qwen2.5-1.5B	Watcher	EQ-DAY-EARN	size-sweep	small	—	—	$0.00 ± $0.00 (passive)	n/a	n/a	n/a
Qwen2.5-3B	Watcher	EQ-DAY-EARN	size-sweep	small	—	—	+$933 ± $1,158	n/a	n/a	n/a
Qwen2.5-7B	Watcher	EQ-DAY-EARN	size-sweep	small	—	—	$0.00 ± $0.00 (passive)	n/a	n/a	n/a
DeepSeek-V4-Pro	Watcher	EQ-DAY-EARN	open-reclaim	open	—	discretion	+$798 ± $798	n/a	n/a	n/a
GLM-5.1	Watcher	EQ-DAY-EARN	open-reclaim	open	—	discretion	+$798 ± $798	n/a	n/a	n/a
Qwen2.5-7B	Watcher	EQ-DAY-EARN	open-reclaim	small	baseline-stale	—	$0.00 ± $0.00 (no action)	n/a	n/a	n/a
claude-fable-5	Strategist	—	reasoning-level	frontier	none	—	$0.00 (degenerate)	n/a	n/a	42.0s
claude-fable-5	Strategist	—	hill-climb	frontier	baseline	—	fitness 82.96 (baseline)	n/a	n/a	n/a
claude-fable-5	Strategist	—	hill-climb	frontier	rv	—	fitness 54.59 (−28.36)	n/a	n/a	n/a
claude-fable-5	Strategist	—	hill-climb	frontier	fb	—	fitness −17.51 (−100.47)	n/a	n/a	n/a
claude-fable-5	Strategist	—	hill-climb	frontier	li	—	fitness −63.48 (−146.44)	n/a	n/a	n/a
claude-fable-5	Strategist	—	hill-climb	frontier	trap	—	fitness −88.51 (−171.46)	n/a	n/a	n/a
claude-fable-5	Strategist	—	hill-climb	frontier	trap+brief	—	fitness −94.48 (−177.44)	n/a	n/a	n/a