Currently listed through these providers:
Model details
Qwen3 235B A22b Thinking 2507
Qwen3 235B A22B Thinking 2507 is a Mixture-of-Experts language model in the Qwen3 family, built around a 235-billion-parameter design where only 22B parameters activate per inference. It is positioned as a "thinking" variant that emphasizes step-by-step reasoning, logical deduction, mathematics, and scientific problem solving rather than general chat. The model extends the Qwen lineage into long-context territory with a 256K token window, making it well suited for academic papers, multi-document analysis, and codebases or research write-ups that exceed shorter context limits. Public artifacts are hosted under the Qwen organization on Hugging Face, and the model is described as achieving state-of-the-art reasoning performance among open-source models at release.
In practical terms, the model fits workflows that need careful, structured reasoning delivered through a chat interface: tutoring and step-by-step explanations, formal verification, scientific QA, and analytical research assistance that may span long documents. The Together AI deployment page lists benchmark scores that frame it against peers, showing GPQA Diamond performance of 80.1% and competitive positions on FrontierMath Tier 4 and related academic tests, useful context for teams comparing reasoning backbones. Because it is open-weight, teams can self-host the model, fine-tune it on domain knowledge, or route it through hosted inference providers, which broadens its fit for both production assistants and experimental reasoning pipelines that need full control over the stack.
Quick Info
Powered by- Provider
- Jiekou.AI
- Model key
- qwen/qwen3-235b-a22b-thinking-2507
- Release date
- Jan 1, 2026
- Last updated
- Jan 1, 2026
- Input modalities
- Output modalities
- Capabilities
Cost
- Input token cost
- $0.30
- Output token cost
- $3.00
Limits
- Output tokens
- 131,072 tokens
- Context window
- 131,072 tokens