Overview
Qwen3.5-122B-A10B is a unified vision-language MoE: 122B total parameters with 10B active, trained with early fusion over multimodal tokens rather than attaching a vision encoder to a text model. Qwen reports it ahead of the Qwen3-VL line across reasoning, coding, agent and visual-understanding benchmarks, with 201 languages and dialects. Served here in FP8 at 262K context, of a documented 1M ceiling.
Strengths
- Strongest visual reasoning in the catalogue
- 262K native context
- 201 languages and dialects
- Only 10B active parameters per token
Use cases
- Dense document, diagram, and chart understanding
- Long video question answering
- Visual agent loops that also need tool calling
Good to know
Thinking content is not surfaced by this deployment: passing enable_thinking=True currently returns empty content, so leave it unset. Roughly three times the cost of qwen3-vl, which is the cheaper choice for bulk image work.
API access
One OpenAI-compatible API for every model in the catalog, plus the CLI and the function SDK inside workflows. Sign up free, add a card for $5 in credits, and these requests work:
curl https://gateway.graphn.ai/v1/chat/completions \
-H "Authorization: Bearer $GRAPHN_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "qwen3.5-122b-a10b-fp8",
"messages": [{"role": "user", "content": "Hello"}]
}'graphn model chat qwen3.5-122b-a10b-fp8 -m "What is GraphN?"
@function
async def ask(question: str) -> str:
return await chat.complete(question, model="qwen3.5-122b-a10b-fp8")