The LeanCTX
community
Every developer running LeanCTX trims the tokens their AI reads — saving money and real energy. This is our shared scoreboard: live, opt-in, summed from tamper-evident local ledgers. Add yours and push the numbers higher.
Collective impact, measured
Updated on every visit — real tokens, real dollars, and the energy that compute never had to burn. Every figure reconciles to the opt-in ledgers below, and grows each time someone new joins.
LiveTokens saved
stripped from LLM context
Money saved
at published model rates
Energy saved
≈ 54,156 phone charges
Fewer tokens means less compute — and less energy burned. Every name on this board is a small, verifiable dent in AI's footprint, multiplied by everyone who joins.
Energy is an estimate — tokens saved × ≈0.4 J/token (modern optimized LLM inference, e.g. Llama-3-70B FP8 on H100); real figures vary by model and hardware. Sources: Epoch AI, John Snow Labs.
Top realized savings
These developers have compressed the most context — and saved the most energy doing it. Claim your spot with a single command; every figure is signed by your local ledger and verifiable offline.
| # | Developer | Tokens saved | Compression | Energy saved | Saved |
|---|---|---|---|---|---|
| 1 | frpboy Leader | 5.1B | 71% | 561.4 kWh | $12,632 est. |
| 2 | yvgude | 574.4M | 71% | 63.8 kWh | $1,436 est. |
| 3 | mas | 79.1M | 70% | 8.8 kWh | $198 est. |
| 4 | cses | 76.2M | 62% | 8.5 kWh | $191 est. |
| 5 | sb | 46.9M | 84% | 5.2 kWh | $117 est. |
| 6 | anonymous | 8.5M | 59% | 947 Wh | $21 est. |
| 7 | Ageratum Studio | 4.5M | 72% | 495 Wh | $11 est. |
| 8 | anonymous | 3.3M | 39% | 369 Wh | $8 est. |
| 9 | webut | 1.4M | 52% | 152 Wh | $3 est. |
| 10 | tana | 814.2K | 64% | 90 Wh | $2 est. |
| 11 | Richard Capraro | 705.6K | 50% | 78 Wh | $2 est. |
| 12 | anonymous | 102.9K | 50% | 11 Wh | $0 est. |
| 13 | Richard Capraro | 30.8K | 70% | 3 Wh | $0 est. |
| 14 | anonymous | 1.7K | 1% | 0 Wh | $0 est. |
Want in? Once LeanCTX has been compressing your
context for a while, run lean-ctx gain --publish --leaderboard — then watch
your name climb.
- Pick your name. Add
--name="your handle"the first time — it sticks for every future push. - No account, ever. Your entry is signed by a key that lives only on your machine. That key is your identity — no login, no email, nothing to leak.
- Push as often as you like. Re-publishing refreshes your one entry in place (same link) — it never piles up duplicates.
- Stay on the board on autopilot.
lean-ctx config set gain.auto_publish truekeeps your numbers fresh as you save more energy.
What's shared: only four aggregate numbers — tokens saved, compression rate, energy and dollars saved — plus the name you pick. Never your code, file paths, repo names or prompts. It's all open source — see the publish contract.
Claim your spot on the leaderboard
Install LeanCTX, let it shrink what your AI reads, and your tokens, dollars and watt-hours start adding to the community total. Publish your recap with one command — your numbers stay verifiable offline, and every saved token is a little less energy burned.