Proof efficiency: the intelligence-multiplier tax.

Since proof v4 every block carries a multiplier from 1× up to 12×. A 1.2× block needs 20% more work than a 1.0× block. Miners and pools that land near 1.0× find more blocks with the same GPUs. This page shows who pays the tax and why.

Network mean, since v4

1.057×

median 1.014× · 90% of blocks ≤ 1.157× · 4,665 blocks

Blocks priced at exactly 1.0×

43%

2,029 of 4,665 v4 blocks

Work spent on the tax

263 blocks

≈ 14,602 TSC ≈ $65,711 at today's price

Rejection line

175 / 256

near-certain steps; highest seen: 165

Who pays the tax

Block finders with ≥3 proof-v4 blocks. "Extra work" = mean multiplier − 1. "Blocks lost" = blocks they would have found at 1.0× with the same work.

Miner / pool	v4 blocks	Mean	p90	Extra work	Blocks lost	Value lost	Late · profile · state	Max near-pin	Main driver
tc1qurhglg…y4wuuv	1,716	1.098×	1.211×	+9.8%	168.2	$42,074	1.018 · 1.078 · 1.000	125 / 175	sampling profile
Tigerpool	1,589	1.029×	1.087×	+2.9%	46.1	$11,529	1.010 · 1.019 · 1.000	120 / 175	occasional spikes
tc1q4w22jf…2p3cla	1,142	1.019×	1.049×	+1.9%	21.7	$5,429	1.008 · 1.011 · 1.000	94 / 175	efficient
tc1qf4uq7y…uewwzy	44	1.011×	1.040×	+1.1%	0.5	$121	1.004 · 1.008 · 1.000	115 / 175	efficient
tc1q2vk4p0…yhtm7l	39	1.030×	1.102×	+3.0%	1.2	$293	1.010 · 1.020 · 1.000	80 / 175	occasional spikes
tc1qg83etp…duejel	26	1.101×	1.188×	+10.1%	2.6	$657	1.000 · 1.101 · 1.000	78 / 175	sampling profile
tc1q54evg5…xlx3k0	24	1.460×	1.604×	+46.0%	11.0	$2,762	1.000 · 1.089 · 1.000	165 / 175	sampling profile
tc1qu64ua8…jpqmwr	22	1.290×	1.598×	+29.0%	6.4	$1,596	1.274 · 1.003 · 1.008	49 / 175	late credit
tc1q3y7r8j…f4a79m	16	1.029×	1.077×	+2.9%	0.5	$116	1.012 · 1.016 · 1.000	95 / 175	occasional spikes
BobLabs	13	1.072×	1.144×	+7.2%	0.9	$234	1.000 · 1.058 · 1.014	102 / 175	sampling profile
tc1q80y237…ykvtwv	10	1.014×	1.048×	+1.4%	0.1	$35	1.000 · 1.014 · 1.000	46 / 175	efficient
tc1qfg6n6p…9h3vaf	3	1.704×	1.552×	+70.4%	2.1	$528	1.428 · 1.012 · 1.178	58 / 175	late credit
tc1q0n7007…hre6mr	3	1.137×	1.000×	+13.7%	0.4	$103	1.137 · 1.000 · 1.000	65 / 175	late credit
tc1qeavkps…7sq054	3	1.297×	1.132×	+29.7%	0.9	$223	1.232 · 1.000 · 1.042	61 / 175	late credit
tc1q4dcqke…qx9f2f	3	1.013×	1.000×	+1.3%	0.0	$10	1.000 · 1.013 · 1.000	72 / 175	efficient

Components are the four factors of the v4 price (the "pin" factor was 1.0 on every block). Pool miners inherit their pool's multiplier: a pool at 1.12× pays everyone ~12% less than the same pool at 1.0× would.

Daily mean multiplier, largest producers

Lower is better. Dashed line = no tax.

tc1qurhglg…y4wuuvTigerpooltc1q4w22jf…2p3clatc1qf4uq7y…uewwzytc1q2vk4p0…yhtm7lideal 1.00×

Distribution

All proof-v4 blocks by multiplier.

1.00× (no tax)up to 1.2×above 1.2×

What moves the multiplier

Late credit: Credit that piles up at the end of the 256-step window or in its last 32–64 steps. Possible causes: custom samplers, unusual prompts, outdated builds. Worth comparing against the reference miner.
Sampling profile: How the proof's token-selection behaviour compares with ordinary inference. Priced above 1.0× when it looks optimised for mining rather than normal generation.
Near-certain steps: Steps where the model is almost sure of the next token. 175 of 256 rejects the block outright.
Reference: The project's honest prompt bank: median 1.07×, 91% of windows ≤ 1.3×.

Source: block headers from the explorer (proof.price_multiplier, proof.price_components, proof.near_pin_count), method from the proof-v4 blog post.