Open source model reports
Reports
When the writing makes a claim about models, there’s a report behind it: five real engineering tasks, the same harness each time, verdicts published whether the local model wins or loses.
-
Qwen3.8-Flash-Next vs Claude Opus 4.7
Qwen3.8-Flash-Next, a 125B-parameter mixture-of-experts with 6B active per token, benchmarked against Claude Opus 4.7 across five real engineering tasks. Local wins test-strategy outright, holds parity on the structured tools, regresses on log-triage, and trails the frontier on commit-description prose. Routing depends on the tool.
-
Qwen 3.8-27B vs Claude Opus 4.7
A local Qwen 3.8-27B benchmarked against Claude Opus 4.7 across five real engineering tasks. Local is adoptable for structured and mechanical work; the frontier model leads on judgement depth.
-
Qwen 3.8-27B vs Claude Opus 4.6
A local Qwen 3.8-27B tested against Claude Opus 4.6, the head-to-head from Qwen's own release benchmarks. Verdict: the vendor positioning holds on three of five tasks. The local model is suitable for structured work; the frontier model leads on prose faithfulness and causal diagnosis.