Observed spend
What developers actually spend on claude-haiku-4-5 — measured from real sessions, not list prices. For specs and pricing tables, look elsewhere; this page only answers what the model costs the people using it.
Broad adoption and low spend per developer — reached for constantly, and cheap enough that it barely registers on the bill.
For scale, spend per developer across tracked models runs from claude-haiku-4-5 at $51 to gpt-5.6-sol at $10,997 — a spread list pricing alone never reveals, because it depends entirely on what people reach for the model to do.
| Model | Per developer | Share |
|---|---|---|
| claude-haiku-4-5 | $51 | 0.4% |
| gpt-5.6-sol | $10,997 | 20% |
| claude-opus-4-8 | $5,387 | 20% |
| claude-fable-5 | $5,099 | 17% |
| gpt-5.5 | $5,275 | 15% |
| claude-opus-5 | $5,892 | 10% |
| claude-opus-4-6 | $1,721 | 7% |
| claude-opus-4-7 | $1,600 | 5% |
| gpt-5.4 | $795 | 2% |
Developers run npx viberank-cli, which reads token counts and API-equivalent costs that ccusage computes from their own local session logs. Spend here is what the usage would cost at model list prices — most people on the board pay a flat subscription instead, so this measures consumption rather than a bill.
Full methodology, caveats and a free JSON endpoint are on the data page. For the distribution across developers rather than models, see what Claude Code actually costs.