---
id: "qwen/qwen3-next-80b-a3b-thinking"
also: ["qwen/qwen3-next-80b-a3b-thinking-2509", "Qwen/Qwen3-Next-80B-A3B-Thinking", "requesty/alibaba/qwen3-next-80b-a3b-thinking", "vercel/alibaba/qwen3-next-80b-a3b-thinking"]
cheapest_list_usd_per_m: "0.15 in, 1.2 out, at alibaba on openrouter"
cheapest_effective_usd_per_m: "0.1494 in, 1.1995 out, at nebius/fp8 on openrouter"
context: 262144
max_output: 235929
input: ["text"]
output: ["text"]
released: "2025-09-11"
params: ["frequency_penalty", "include_reasoning", "logit_bias", "logprobs", "max_tokens", "presence_penalty", "reasoning", "repetition_penalty", "response_format", "seed", "stop", "structured_outputs", "temperature", "tool_choice", "tools", "top_k", "top_logprobs", "top_p"]
endpoints: 6
gateways: "openrouter (3), vercel (2), requesty (1)"
no_training: 6
trains: 0
retains_prompts: 1
intelligence: 11.2000
coding: 17.4000
---
# Qwen: Qwen3 Next 80B A3B Thinking

Qwen3-Next uses a highly sparse MoE design: 80B total parameters, but only \~3B activated per inference step. Experiments show that, with global load balancing, increasing total expert parameters while keeping activated experts fixed steadily reduces training loss.Compared to Qwen3’s MoE (128 total experts, 8 routed), Qwen3-Next expands to 512 total experts, combining 10 routed experts + 1 shared expert — maximizing resource usage without hurting performance. The Qwen3-Next-80B-A3B-Thinking excels at complex reasoning tasks — outperforming higher-cost models like Qwen3-30B-A3B-Thinking-2507 and Qwen3-32B-Thinking, outpeforming the closed-source Gemini-2.5-Flash-Thinking on multiple benchmarks, and approaching the performance of our top-tier model Qwen3-235B-A22B-Thinking-2507.