Comparisons

Four AI blog subscriptions priced from their own pages, and one published post from each run through our house checks. The prose passes. Nine statistics between them, and not one points at a source outside the industry.

One inbox, eight days, 144 threads and exactly one message from a person. We priced Gemini, Copilot, Shortwave, Fyxer and Missive against what was actually there.

Seven models sit above 95% on SWE-bench Verified. On repositories picked so they were unlikely to have trained on them, the best score is 61.5%. We read four leaderboards, one controlled trial and three vendor pricing pages.

Seat licence, token meter or rented GPU: three ways to buy an artificial intelligence platform, priced against one defined workload from the vendors’ own pages.

Four different products are sold as generative AI platforms: consumer seats, first-party model APIs, cloud marketplaces and open weights. We priced all four on the same workload from vendor pages read on 18 August 2026.

We asked three fixed questions of six free AI tiers: what they cost in money, in throughput, and in what the vendor keeps of whatever you type.

Three coding agents finish within 3.4 points on the same benchmark, entry pricing has converged on $10 to $20 a month, and the best controlled study still cannot separate the effect from zero.

Nobody has ever failed to build an agent because the loop was hard. They fail on the retries, budgets and logging a demo never needed.

A stated window is a capacity figure, not a usable-attention figure. The middle of a long input is measurably a worse place to be.

The API bill is visible on an invoice. The self-hosting bill is spread across salaries, on-call rotas and slipped projects.

A second datastore is a second thing to back up, monitor, secure and be woken by. That cost is invisible in a benchmark.

Serving one person and serving two hundred are different engineering problems. The tool that wins at one loses badly at the other.