Qwen3.8-Flash-Next Beats Claude Opus on SWE-bench Pro at 1/9 Cost
eyishazyer · x · 2026-08-28
Qwen3.8-Flash-Next claims to top the SWE-bench Pro leaderboard with a score of 62.5, surpassing Claude Opus 4.6 Max (53.4). It is an open-weight multimodal MoE with 125B total parameters but only 6B active per token. Crucially, Qwen states it was trained for roughly 1/9 the cost of its predecessor, Qwen3.7-Plus, which it now outperforms on most coding tasks.
Key Metrics:
- SWE-bench Pro: 62.5
- LiveCodeBench: 91.9
- GPQA Diamond: 91.7
- Context: 262K native, extendable to 1M.
Caveats:
- Claude Opus 4.6 Max still leads on the HLE benchmark (40.0 vs 35.9).
- There are inconsistencies in Qwen's own documentation regarding 1M context speedup claims (marketing cites 7.6x/4.9x, while cookbooks cite 10.2x/6.6x). Independent benchmarking is needed.
More from coding & agent
- Animated teardown of XBOW's Chrome full-chain exploit that won Google's first $250K VRP bounty — moyix · 2026-10-15
- pstack's durable core: 10-20 line skills that distill agent best practices into principles — HankYeomans · 2026-10-15
- 'Children of Elon' forum lets AI agents post and teach each other while humans watch — AiJohnAllen · 2026-10-15
- VoidZero open-sources Void: SDK, CLI and platform to deploy Vite apps on Cloudflare — cnakazawa · 2026-10-15
- Running 2 agents on the same data breaks everything — Omnigraph offers a fix — sarahdrinkwater · 2026-10-15
- Evan You presents 'Developer Tooling in the Agentic Era' at ViteConf 2026 — cnakazawa · 2026-10-15