Qwen3.5 35B A3B model achieves high GPQA score and cost-efficiency

The Qwen3.5 35B A3B model has demonstrated impressive performance, achieving an 81.9% score on the GPQA benchmark. It also offers a high intelligence-to-cost ratio, delivering 35.3 intelligence points per dollar. The model's speed is notable, processing at 166.7 tokens per second, with its overall performance exceeding expectations for its parameter count. AI

IMPACT Demonstrates significant performance gains relative to parameter count and cost, potentially influencing future model development and adoption.

RANK_REASON The item reports on a specific benchmark performance for an AI model. [lever_c_demoted from research: ic=1 ai=1.0]

Read on Mastodon — mastodon.social →

• GPQA: A Graduate-Level Google-Proof Q&A Benchmark
• Qwen3.5 35B A3B

• model release
• paper

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →