MON, AUGUST 03, 2026
Independent · In‑Depth · Practitioner‑Tested
Large Language Models

Qwen3.8-Max-Preview vs DeepSeek V4 Flash 0731 (2026): Large Unverified vs Small Verified

2.4T Preview With No Benchmarks vs 284B With Terminal-Bench 82.7 at $0.14/M

🕐 5 min read 👁 14 views 📅 Aug 3, 2026

QUICK VERDICT — AUGUST 2026

V4 Flash 0731 for agentic coding: $0.14/M, Terminal-Bench 82.7%, Codex-compatible, MIT weights. The verified cost-performance leader for agentic coding workloads today. Migrate from deepseek-chat before October 24, 2026.

Qwen3.8-Max-Preview for multimodal evaluation: Test via Token Plan for document-heavy, image, or general frontier tasks. Not ready for production until Alibaba publishes benchmarks, per-token pricing, and open-weight checkpoint.

Last updated August 3, 2026. Related: Qwen3.8-Max full fact-check → · DeepSeek V4 Flash 0731 review →

⚖ Our Verdict

DeepSeek V4 Flash 0731 wins on every verifiable dimension today: Terminal-Bench 82.7%, $0.14/M confirmed, MIT open weights live. Qwen3.8 likely exceeds V4 Flash on raw intelligence and multimodal capability — unverifiable without published benchmarks. Different primary use cases: V4 Flash is an agentic coding specialist; Qwen3.8 is a general multimodal frontier model.