How TreadLightlyAI compares
Side-by-side with the major chatbots on privacy, sustainability, and transparency, so you can decide what fits.
Last updated 2026-08-03
Models in TreadLightlyAI
Five chat tiers, four engines, three providers, with blended energy rates per call. See our energy methodology for how every rate below was derived.
| Model | Creator | Underlying Model | Parameters | Energy Rate | Data Center | Green Energy | Verified |
|---|---|---|---|---|---|---|---|
| Green Tiny | Google DeepMind | Gemma 4 31B Instruct (capped output) | 31B | ~108 µWh/token(estimated) | Infomaniak (Geneva, CH) | Yes | 2026-08-03 |
| Green Small | Google DeepMind | Gemma 4 31B Instruct | 31B | ~108 µWh/token(estimated) | Infomaniak (Geneva, CH) | Yes | 2026-08-03 |
| Green Large | OpenAI | GPT-OSS 120B | 120B | ~115 µWh/token(measured) | GreenPT / Scaleway (EU) | Yes | 2026-04-20 |
| Haiku | Anthropic | Claude Haiku 4.5 | Not disclosed | ~78 µWh/token(estimated) | Anthropic (US) | No | 2026-04-20 |
| Sonnet | Anthropic | Claude Sonnet 5 | Not disclosed | ~144 µWh/token(estimated) | Anthropic (US) | No | 2026-07-17 |
Energy rates are blended (input = output). Green Large's rate is measured from GreenPT per-call data; every other rate, including Green Tiny and Green Small, is estimated (methodology version ). Green Tiny and Green Small share one engine (Gemma 4 31B at Infomaniak) and carry the same rate. Worth saying plainly: our estimate for Claude Haiku is lower than our estimate for Green Tiny and Green Small. The green tiers are chosen for where their energy comes from and who trained the model, not for using less of it than everything else on offer. Full methodology
Transparency commitments
What each service publishes about its data handling, energy use, and model provenance.
| Dimension | TreadLightlyAI | ChatGPT(opens in new tab) | Gemini(opens in new tab) | Copilot(opens in new tab) | Claude(opens in new tab) |
|---|---|---|---|---|---|
| Real-time energy tracking | Yes | No | No | No | No |
| Anti-sycophancy by design | Yes | Partial | No | No | Partial |
| Donates to AI safety/environment | 10% to FLI, 1.5% carbon removal | No | No | No | Partial |
| Published AI limitations | AI Transparency Card | Partial | Partial | Partial | Yes |
| No user data training | Yes | No | No | Partial | Partial |
| Financial transparency | Partial | No | Partial | Partial | No |
| AI alternatives guidance | Yes | No | No | No | No |
Feature Comparison
Beyond our transparency commitments, here's how TreadLightlyAI compares as a product. Every item below reflects what's live today: no roadmap padding, no asterisks.
| Dimension | TreadLightlyAI | ChatGPT(opens in new tab) | Gemini(opens in new tab) | Copilot(opens in new tab) | Claude(opens in new tab) |
|---|---|---|---|---|---|
| Large Frontier Models | No (intentional) | Yes (flagged)GPT-5.4 | Yes (flagged)Gemini 2.5 Pro | Yes (flagged)GPT-5.4 | Yes (flagged)Opus |
| Medium Models | Green Large (GPT-OSS) | GPT-4o-mini | Gemini 2.0 Flash | GPT-4o-mini | Sonnet, Haiku |
| Efficient Models | Green Small, Green Tiny | No | Flash Lite | No | No |
| Code assistance | Basic | Yes | Yes | Yes | Yes |
| Thinking modes | Yes | Partial | Partial | Partial | Partial |
| Prompt improvement | Hints and tools | No | No | No | No |
| Project Folders | Yes | Yes | Gems | Partial | Yes |
| Conversation history & export | Yes | Yes | Yes | Partial | Partial |
| Image & document analysis | All tiers, green models included | Yes | Yes | Yes | Yes |
| Image generation | No (intentional) | Yes (flagged) | Yes (flagged) | Yes (flagged) | No (intentional) |
| Web search | Yes | Yes | Yes | Yes | Yes |
| Enterprise features | No | Yes | Yes | Yes | Yes |
| Free tier | Yes | Yes | Yes | Yes | Yes |
| Pricing (paid tiers) | $3, 8, 20 /mo | $20, 200 /mo | $20/mo | $20/mo | $20, 100, 200 /mo |
If TreadLightlyAI is not a fit, you have other good choices.
Responsible AI is bigger than us. We maintain a comparison of responsible AI chatbots at Dangerous Robot: Responsible AI(opens in new tab)