Directory
Compare models vs apps — live OpenRouter pricing, no fake benchmarks
Pricing moves this week
| Model | Tier | Best for | Input | Context | Status |
|---|---|---|---|---|---|
| DeepSeek (via Ollama)DeepSeek | Local | Best for strong open-weight reasoning/coding locally; weak if you lack VRAM for mid-size tags. | — | — | AVAILABLE |
| Llama (via Ollama)Meta | Local | Llama is best used for natural language processing tasks and conversational AI applications. | — | — | AVAILABLE |
| Mistral (via Ollama)Mistral AI | Local | Mistral is best used for high-performance natural language processing tasks. | — | — | AVAILABLE |
| Stable Diffusion (via ComfyUI)Stability AI / community | Local | Best for unlimited local image pipelines; weak if you want one-click cloud simplicity. | — | — | AVAILABLE |
| LM StudioCommunity | Local | Best for a GUI on-ramp to local GGUF models; weak if you expected a model (it's an app). | — | — | AVAILABLE |
Free tokens & offers
Verified free entry points — terms change.
Free models, one API key
Dozens of models listed at $0 for both input and output tokens — rate-limited but genuinely free. Any OpenRouter-compatible client works.
Verified 2026-09 · terms may change
Gemini API free tier
Free-of-charge tier for the Gemini API with generous per-minute quotas — a card is required for billing identity, but the free quota isn't billed.
Verified 2026-09 · terms may change
Fast inference, free tier
Free tier on GroqCloud for Llama, Mixtral and other open models at very high tokens/second — great for prototyping agents and chat apps.
Verified 2026-09 · terms may change
La Plateforme experiment tier
Experiment plan lets you try Mistral models via API for free within a small usage envelope before committing to a paid plan.
Verified 2026-09 · terms may change
GitHub Models playground
Free API access to GPT, Llama, Mistral and more for GitHub users within daily rate limits, straight from your GitHub account.
Verified 2026-09 · terms may change
Free tier on wafer-scale inference
Free tier for extremely fast inference of open models — useful when latency, not cost, is your bottleneck.
Verified 2026-09 · terms may change