Network work rate and who finds the blocks.

Mining models

Since block #31,528 more than one model produces blocks. All models share one difficulty adjustment, but each block's difficulty is written in its own model's units: a Qwen3.8-27B block carries about 20× less difficulty than a Qwen3-8B block at the same moment. This site converts every block to Qwen3-8B-equivalent work, so work rate, earnings and cost add up across models.

Model per block from the explorer (field model). Weight = median ratio of base difficulty between neighbouring blocks of the two models, measured on-chain. Mining calculator and GPU rates are for Qwen3-8B: no public per-GPU rates for Qwen3.8-27B yet.

Progress through epoch 6: 43.4% · date assumes the 7-day average block time (9.9 min). If the network keeps the same work rate, each miner's TSC per day falls by 40% at the cut.

Network work rate

Combined work of all miners in proof/s (sum of effective block difficulty ÷ time). Blocks of other models are converted to Qwen3-8B-equivalent work, see Mining models. The hourly estimate is inherently noisy — at ~6 blocks per hour, luck dominates; the 24-hour average is the reliable line.

What does it cost to mine 1 TSC?

GPU running cost ÷ the TSC that GPU earns at the current network work rate. A GPU doing R PoI/s on a network doing N PoI/s earns R/N of all new TSC. Rising network work rate and falling block rewards push this cost up; if the price falls below it, miners with that setup mine at a loss.

Cost to mine 1 TSC vs. price, per UTC day

History starts 20 Jul 2026, the first day the network exceeded 1,000 PoI/s. Days before the SafeTrade listing have no price line.

By GPU, last 24 hours

Estimates, not quotes. Power uses the vendor board limit (TDP); real draw during inference is often lower, so electricity cost is a conservative upper bound. Not included: pool fee (~1%), hardware depreciation, cooling, luck. PoI/s from tsc.suprnova.cc (Start Mining → rates table), 29 Sep 2026; rental from cheapest spot listings on Vast.ai, Sep 2026.

Who finds the blocks?

Every block records the payout address that produced it. Over many blocks, share of blocks ≈ share of work. Red line = 50%: one pool above it could in principle rewrite recent blocks; two pools together only by coordinating.

Work rate per miner = sum of effective difficulty of its blocks in 24h ÷ 86,400 s, in Qwen3-8B-equivalent units. Over one day, ±3–5 points of share is ordinary luck. Pool aliases come from the explorer; unlabelled addresses are probably solo miners or unlabelled pools.

Share of blocks per UTC day

Last 45 days, top 5 addresses of the past 7 days. Today is a partial day.

What does mining earn?

What 1 K proof/s would earn running for a whole UTC day at that day's difficulty, before pool fees, electricity and luck. Multiply by your miner's speed. USD uses that day's SafeTrade close — it shows scale, not income anyone can count on.

The reward falls when more miners join and rises when they leave; the second driver is the shrinking block reward (epochs).

Proof-of-inference — work multiplier

In TSC a block can count as more work than its bits imply if the miner attached a “useful” inference proof (multiplier > 1). A high average multiplier means miners are actually running inference, not just hashing.