Speed
| What is measured | Value | What it means | Source · read on |
|---|---|---|---|
| Market reply latency, median | 0.86 ms | Tokyo host; 248.6 ms from our other host (×289). | Published report, 27 Oct 2026 · 2026-10-27 |
| REST time to first byte, median | 35 ms | Tokyo host; 1 001 ms from our other host (×28). | Published report, 27 Oct 2026 · 2026-10-27 |
| Event throughput, Rust core | 33 973 events/s | Same machine, same capture: Go core 19 956 events/s (+70%). | Published report, 27 Oct 2026 · 2026-10-27 |
| Core tail latency, p99 | 64 µs | Go core on the same machine: 123 µs. | Published report, 27 Oct 2026 · 2026-10-27 |
Price of a run
| What is measured | Value | What it means | Source · read on |
|---|---|---|---|
| A four-line review of one proposal | 0.0086 USD | Two metered lines; the other two spend plan quota, which is not free. | Council receipts, 27 Oct 2026 · 2026-10-27 |
| One cheap line, one full answer | 0.00601019 USD | 318 620 tokens, 116 s of wall time. | Transport receipt, 27 Oct 2026 · 2026-10-27 |
| A frontier judge run, Gemini 4 Argon | 0.0607607 USD | Provider-reported cost of one verdict, 22:15:13 UTC. | Verdict header, 27 Oct 2026 · 2026-10-27 |
| A frontier judge run, Grok 4.7 | 0.07049 USD | Provider-reported cost of one verdict, 22:16:50 UTC. | Verdict header, 27 Oct 2026 · 2026-10-27 |
| One image generation in the studio | 0.25 USD | Reserve 0.06 USD, daily cap 1 USD; the charge is fixed after the provider answers. | Studio capture receipt, 27 Oct 2026 · 2026-10-27 |
Efficiency
| What is measured | Value | What it means | Source · read on |
|---|---|---|---|
| Tokens inside one cheap answer | 318 620 | 14 API calls inside that run, not a single message. | Transport receipt, 27 Oct 2026 · 2026-10-27 |
| Paper trades from 927 cohorts | 0 | 368.9 million frames; no coin passed the entry gate. Published as a negative result. | Tokyo report, 27 Oct 2026 · 2026-10-27 |
| Coin observations, none unblocked | 13 549 | The gate, not the market silence, stopped every candidate. | Tokyo report, 27 Oct 2026 · 2026-10-27 |
Our data
| What is measured | Value | What it means | Source · read on |
|---|---|---|---|
| Commits landed, 24–27 October | 67 | Count as of 00:10 UTC on 28 October; activity, not value. | Repository history · 2026-10-28 |
| Acceptance criteria passed by the browser demo | 18 / 18 | Every criterion on our own list passed at 12:17:50 UTC. | Acceptance receipt · 2026-10-27 |
| Built-in files matching the public package | 71 | Checked in a Windows smoke run against the anonymous download. | Release journal · 2026-10-27 |
| Addresses in the site map | 462 | Across English, Russian, Ukrainian and eight more locales. | Traffic report · 2026-10-27 |
| Sessions in the last 28 days | 57 | 15 users, 549 page views, 50 engaged sessions. | Traffic report · 2026-10-27 |
| Users in the last 28 days | 15 | Search Console rows are still pending; we write that as pending, not as zero. | Traffic report · 2026-10-27 |
| Frames replayed in the paper experiment | 368.9 million | 927 cohorts, 13 days, zero trades. | Tokyo report · 2026-10-27 |
| Facts behind the articles, each with a source | 21 | The claim gate refuses any number that is not in this list. | Evidence base, 27 Oct 2026 · 2026-10-27 |
Models
| What is measured | Value | What it means | Source · read on |
|---|---|---|---|
| Model lines that answered one question | 4 of 5 | The fifth line exists and is sold, but our launcher had no entry for it — our defect. | Council report, 27 Oct 2026 · 2026-10-27 |
| Votes on the accepted items | 4/4 · 3/3 · 4/4 | Single version source, first-user path, diff inside the receipt. | Council report, 27 Oct 2026 · 2026-10-27 |
| Sakana Fugu on our tasks | not measured | Vendor description only; we did not reproduce it and quote no price. | Verdict, 27 Oct 2026 · 2026-10-27 |
Need a run like this for your own task?Send a request →
Filters never change a value; they only hide rows. If a number here disagrees with another page, the source named in the row wins and the other page is wrong.