Currently listed through these providers:
Model details
Qwen3.6 35B-A3B
Qwen3.6 35B A3B is a multimodal reasoning model that accepts text, image, and video inputs and returns text outputs, sized for workflows that mix long documents with visual or video context. The "A3B" naming hints at an active-parameter design typical of Qwen's recent MoE-style lineup, though specific architecture and parameter counts are not detailed in the supplied sources. It is also cataloged on NVIDIA NGC as a NIM entry under Qwen's team, suggesting it can be deployed through NVIDIA's inference stack in addition to third-party gateways. Practical use cases lean toward agentic and tool-driven tasks: the model exposes a broad parameter surface that includes reasoning controls, temperature, top-p and top-k sampling, structured outputs, response formats, tools and tool choice, repetition penalty, and seed, which is a strong fit for production pipelines that need controllable generation and reproducible runs. The 262,144-token context window allows it to keep very large inputs in view at once, useful for long-document analysis, multi-image review, or extended video-conditioned reasoning. Routing through a gateway model layer adds flexibility for teams standardizing on OpenRouter-style access patterns, while keeping the option to run the same model on NVIDIA NIM infrastructure for on-prem or higher-throughput deployments.
Because the evidence comes mainly from a single third-party integration guide that mirrors OpenRouter metadata, any claims about training data, benchmark scores, or licensing status are not confirmed and should be treated as unverified. The guide lists reasoning and temperature among supported parameters, which aligns with the model's intended use as a controllable reasoning engine rather than a raw chat baseline. For evaluators, the practical signals are the multimodal input coverage, the very large context window, the rich sampling and tool-use controls, and the dual availability through both a hosted gateway and NVIDIA's NGC catalog, all of which point to a model designed to slot into complex, long-context, agentic applications.
Quick Info
Powered by- Provider
- Kilo Gateway
- Model key
- qwen/qwen3.6-35b-a3b
- Release date
- Apr 17, 2026
- Last updated
- Apr 17, 2026
- Input modalities
- Output modalities
- Capabilities
Cost
- Input token cost
- $0.10
- Output token cost
- $0.90
Limits
- Output tokens
- 235,929 tokens
- Context window
- 262,144 tokens