Think Deeper, Act Faster: Qwen3-235B-A22B-Thinking-2507 Now Available on SiliconFlow

With Qwen3-235B-A22B-Instruct-2507 (Non-Thinking mode) already demonstrating exceptional performance on SiliconFlow, today we're excited to bring the next breakthrough to our model catalog: Qwen3-235B-A22B-Thinking-2507. This newly open-source model delivers exceptional advances in both reasoning performance and general intelligence, matching the capabilities of leading proprietary models such as Gemini-2.5 Pro and O4-mini while establishing new performance benchmarks for open-source AI.

From advanced research analysis to complex code generation, developers now have access to unprecedented reasoning performance for sophisticated problem-solving tasks.

With SiliconFlow's Qwen3-235B-A22B-Thinking-2507 API, you can expect:

• Cost-Effective Pricing: $0.35/M tokens (input) and $1.42/M tokens (output).

• Extended Context Window: 256K context window for complex tasks.

Key Capabilities & Benchmark Performance

Compared to previous open-source models like DeepSeek-R1-0528, Qwen3-235B-A22B-Thinking-2507 demonstrates significant improvements in practical capabilities:

• SOTA Reasoning Performance: Significantly improved logical reasoning, mathematics, science, coding, and academic benchmarks that typically require human expertise — achieve state-of-the-art results among open-source thinking models.

• Enhanced General Capabilities: Better instruction following, tool usage, text generation, and alignment with human preferences.

• Extended Long-context Understanding: Enhanced 256K long-context understanding capabilities.

These capabilities are reflected in the model’s strong and balanced performance across multiple industry-standard benchmarks.

It ranks first among all compared models on LiveCodeBench v6 and Arena-Hard v2, demonstrating superior coding ability and alignment with human preferences. On AIME25, it achieves 92.3 — outperforming Gemini-2.5 Pro (88.0) and matching O4-mini (92.7)—showcasing advanced mathematical reasoning.

These results demonstrate that Qwen3-235B-A22B-Thinking-2507 is one of the most capable open-source models to date, with competitive performance even against leading proprietary systems.

Real Application Scenarios

Available now on SiliconFlow, Qwen3-235B-A22B-Thinking-2507 features enhanced thinking capabilities with long-context understanding.

• Healthcare Intelligence

Medical literature analysis, clinical decision support, and precision medicine insights derived from patient records and research databases. Analysis of genetic variations, drug interactions and treatment protocols. Perfect for diagnostic assistance, research evidence synthesis, and personalized treatment planning.

• Educational Enhancement

Interactive tutoring in complex STEM subjects, programming instruction and personalized learning design. Adapts explanations and step-by-step guidance to individual learning styles and cognitive needs. Ideal for advanced mathematics, coding bootcamps, and research methodology training.

• Business Document Intelligence

Document analysis across contracts, technical specifications, and regulatory filings with contextual cross-referencing. Extracts key insights, identifies compliance risks, and generates executive summaries. Suited for legal document review, due diligence, and knowledge management systems.

Get Started Immediately

• Explore: Try Qwen3-235B-A22B-Thinking-2507 in the SiliconFlow playground.

• Integrate: Use our OpenAI-compatible API. Explore the full API specifications in the SiliconFlow API documentation.

import requests

url = "https://api.siliconflow.com/v1/chat/completions"

payload = {
    "model": "Qwen/Qwen3-235B-A22B-Thinking-2507",
    "messages": [
        {
            "role": "user",
            "content": "Tell me a story"
        }
    ]
}
headers = {
    "Authorization": "Bearer <token>",
    "Content-Type": "application/json"
}

response = requests.request("POST", url, json=payload, headers=headers)

print(response.text)

Ready to unlock advanced reasoning capabilities?

Explore Qwen3-235B-A22B-Thinking-2507 on SiliconFlow today.