Gemma 4 26B A4B IT is an instruction-tuned Mixture-of-Experts (MoE) model from Google DeepMind. Despite 25.2B total parameters, only 3.8B activate per token dur
No signup required — 10,000 session credits for guests. 50,000 free credits on account creation.
Gemma 4 26B A4B IT is an instruction-tuned Mixture-of-Experts (MoE) model from Google DeepMind. Despite 25.2B total parameters, only 3.8B activate per token during inference — delivering near-31B quality at...
Very large context window — 262,144 tokens (paste entire codebases or long PDFs).
Free tier at $0 upstream cost — great for experimenting without watching the meter.
Hosted by Google — you can access it here alongside GPT-4o, Claude, Gemini and 100+ more on one plan.
Switch to any other model mid-conversation from the picker.
Free-tier on the upstream side — on Faceb.ai we still charge a minimum 1 credit per call to cover transit, so effectively unlimited on the typical session.
Google: Gemma 4 26B A4B (free) is a chat/completion model served by Google and accessed through Faceb.ai. Try it without signing up — guest users get 10,000 session credits to experiment.
Yes — this model is on the free-tier of its provider, so usage is effectively free (rate-limited upstream). Faceb.ai still deducts a minimum 1 credit per call.
262,144 tokens. Paste your source material in, no need to truncate.
Yes. Any API key from /account/api/ works with model slug `google/gemma-4-26b-a4b-it:free`. The OpenAI SDK works with base_url=https://api.faceb.ai/v1.
Depends on the task. Faceb.ai lets you switch models per message — benchmark side-by-side by asking both the same prompt, which is more reliable than abstract comparisons.
Free-tier on the upstream side — on Faceb.ai we still charge a minimum 1 credit per call to cover transit, so effectively unlimited on the typical session.
No. We contractually request that upstream providers not train on content routed through us. Your chat history lives only on your account.
It depends on the model size and training mix. For serious code work the developer favourites are Claude 3.5 Sonnet and DeepSeek V3; for quick edits, most capable models work fine.
Yes — point your SDK at https://api.faceb.ai/v1 and use this model's slug as the model parameter. Everything else works as normal.
Check the model catalog — multimodal models are marked in the picker. If it accepts images, you can drop screenshots and diagrams straight into the chat.
Yes — the picker is always at the top of the chat. Previous context carries over.
Yes. Our catalog auto-fetches from the upstream aggregator, so provider updates and new versions appear in the picker as soon as they're available.
Your Faceb.ai credits work for every model — switch per message, no extra subscriptions.
OpenAI's flagship multimodal model — text, vision, and code.
Chat with GPT-4o →OpenAI's fastest, cheapest frontier model — great default.
Chat with GPT-4o mini →Anthropic's best balance of quality and cost — a coder favourite.
Chat with Claude 3.5 Sonnet →One subscription covers every frontier model — switch between them per message. No extra API keys, no extra bills.
Start chatting with Google: Gemma 4 26B A4B (free) → Go Pro · $14.99/mo