Numbers need a cohort before they need a rank.
A real model creates a file and edits it in the same session. The result is checked outside the session: pages in headless Chrome, scripts by running them. 138 sessions, medians.
The gain comes from the generator. On a task no generator covers, Simplicio can cost more than no MCP at all, and the chart shows it.
| Model | Task | Setup | n | Time | Model requests | Output tokens | Cost | Delivered |
|---|---|---|---|---|---|---|---|---|
| Claude Sonnet 5.5 · Claude Code | Page from a JSON file, a generator covers it | Simplicio 3.10.8 | 5 | 17.5 s | 8 | 1,270 | $0.0672 | 5/5 |
| No MCP | 5 | 33.4 s | 5 | 5,147 | $0.0927 | 5/5 | ||
| New page, other data, other language, not used to build the generator | Simplicio 3.10.8 | 5 | 16.3 s | 8 | 1,084 | $0.0514 | 5/5 | |
| No MCP | 5 | 22.5 s | 5 | 2,819 | $0.0610 | 5/5 | ||
| Python script, no generator covers it | Simplicio 3.10.8 | 5 | 19.6 s | 7 | 1,582 | $0.0548 | 5/5 | |
| No MCP | 5 | 11.4 s | 5 | 644 | $0.0283 | 5/5 | ||
| Grok 4.7 · Grok Build | Page from a JSON file, a generator covers it | Simplicio 3.10.8 | 3 | 31.0 s | 8 | 1,289 | $0.1690 | 3/3 |
| No MCP | 3 | 359.7 s | 19 | 28,247 | $0.8590 | 3/3 | ||
| New page, other data, other language, not used to build the generator | Simplicio 3.10.8 | 3 | 36.0 s | 8 | 1,598 | $0.2059 | 3/3 | |
| No MCP | 3 | 387.5 s | 22 | 27,598 | $0.8518 | 3/3 | ||
| Python script, no generator covers it | Simplicio 3.10.8 | 3 | 82.0 s | 11 | 5,510 | $0.3315 | 3/3 | |
| No MCP | 3 | 63.1 s | 9 | 4,009 | $0.2452 | 3/3 | ||
| Claude Haiku 5.5 · OpenCode | Page from a JSON file, a generator covers it | Simplicio 3.10.8 | 5 | 13.7 s | 6 | 918 | $0.0046 | 5/5 |
| No MCP | 5 | 29.3 s | 5 | 6,917 | $0.0083 | 5/5 | ||
| New page, other data, other language, not used to build the generator | Simplicio 3.10.8 | 5 | 13.8 s | 6 | 794 | $0.0023 | 5/5 | |
| No MCP | 5 | 28.7 s | 5 | 6,310 | $0.0057 | 5/5 | ||
| Python script, no generator covers it | Simplicio 3.10.8 | 5 | 19.2 s | 7 | 1,583 | $0.0033 | 5/5 | |
| No MCP | 5 | 16.6 s | 7 | 1,358 | $0.0030 | 5/5 | ||
| DeepSeek V4.1 Flash · OpenCode | Page from a JSON file, a generator covers it | Simplicio 3.10.8 | 10 | 14.1 s | 6 | 602 | $0.0021 | 10/10 |
| No MCP | 10 | 176.3 s | 32.5 | 17,929 | $0.0203 | 10/10 | ||
| Simplicio 3.9.2 | 10 | 166.2 s | 24.5 | 19,944 | $0.0215 | 10/10 | ||
| New page, other data, other language, not used to build the generator | Simplicio 3.10.8 | 6 | 14.1 s | 6 | 524 | $0.0010 | 6/6 | |
| Simplicio 3.9.2 | 6 | 110.9 s | 17.5 | 16,091 | $0.0153 | 6/6 | ||
| Python script, no generator covers it | Simplicio 3.10.8 | 6 | 30.7 s | 10.5 | 1,796 | $0.0031 | 6/6 | |
| No MCP | 6 | 22.7 s | 8 | 1,268 | $0.0024 | 6/6 | ||
| Simplicio 3.9.2 | 6 | 34.1 s | 11.5 | 2,485 | $0.0051 | 6/6 |
Each model runs in one fixed setup, the same on both sides: Claude Code, Grok Build, or OpenCode through OpenRouter. Cost is the list price of the tokens used. These are small tasks: other tasks will give other numbers.
| Metric | Requirement for claim | Current state | |||||
|---|---|---|---|---|---|---|---|
| Ranking against other tools | 10 | 166.2 s | 24.5 | 19,944 | $0.0215 | 10/10 | |
| No MCP | 10 | 176.3 s | 32.5 | 17,929 | $0.0203 | 10/10 | |
| New page, other data, other language, not used to build the generator | Simplicio 3.10.8 | 6 | 14.1 s | 6 | 524 | $0.0010 | 6/6 |
| Simplicio 3.9.2 | 6 | 110.9 s | 17.5 | 16,091 | $0.0153 | 6/6 | |
| Python script, no generator covers it | Simplicio 3.10.8 | 6 | 30.7 s | 10.5 | 1,796 | $0.0031 | 6/6 |
| Simplicio 3.9.2 | 6 | 34.1 s | 11.5 | 2,485 | $0.0051 | 6/6 | |
| No MCP | 6 | 22.7 s | 8 | 1,268 | $0.0024 | 6/6 |
OpenCode 2.1.4 and DeepSeek V4.1 Flash on one fixed OpenRouter endpoint, the same for every run. Cost is computed at $0.15 input, $0.60 output and $0.003 cache read per million tokens. This is one model on three small tasks: other tasks will give other numbers.
| Metric | Requirement for claim | Current state |
|---|---|---|
| Ranking against other tools | Same task, model, hardware, n and quality threshold | N/A |
| Token economy | 1 - tokens_with/tokens_baseline, with cohort and provider price | Measured |
| Route smoke | Versioned static build and local HTTP check | Measured locally |