Now Available: Qwen3.5-35B-A3B and Mercury 2 — Efficient MoE and Frontier Reasoning

Two more models from this week's releases are live on the Hanzo AI Gateway.

Qwen3.5-35B-A3B — Multimodal MoE, Apache 2.0

Released February 24 by the Alibaba Qwen team, this is a 35B-parameter MoE model with only 3B active parameters per forward pass. It's natively multimodal (text + image input) and ships under the Apache 2.0 license — fully permissive for commercial use.

Key specs:

• 35B total / 3B active — extremely efficient MoE
• Multimodal — text and image input, text output
• Apache 2.0 — no restrictions on commercial use
• Outperforms models 6x its size on standard benchmarks

For teams running agents at scale, a model with 3B active parameters that matches the quality of much larger dense models is a game-changer for inference costs.

curl https://llm.hanzo.ai/v1/chat/completions \
  -H "Authorization: Bearer $HANZO_API_KEY" \
  -d '{"model": "qwen/qwen3.5-35b-a3b", "messages": [{"role": "user", "content": "..."}]}'

Mercury 2

Also released February 24, Inception's Mercury 2 brings frontier reasoning capabilities in a proprietary model format. Available now through the gateway.

curl https://llm.hanzo.ai/v1/chat/completions \
  -H "Authorization: Bearer $HANZO_API_KEY" \
  -d '{"model": "inception/mercury", "messages": [{"role": "user", "content": "..."}]}'

100+ Models, One API

With these additions, the Hanzo AI Gateway now serves well over 100 models from every major provider. Every model — open or proprietary — is available through the same endpoint, same API key, same billing. Zero markup on third-party pricing.

Browse the full catalog and start building at hanzo.ai.