# Qwen 3.6 Plus on the Cotera agent benchmark

> 0/5. Burned $1.91 on Crunchbase alone to produce zero parseable output. The worst run in the matrix.

Source: https://cotera.co/benchmarks/models/fw-qwen3p6-plus

---

Alibaba's Qwen 3.6 Plus — the largest dense Qwen variant in the November 2026 lineup, pitched as a frontier reasoning option from the Chinese open-weights ecosystem.

**Score:** 0/5 benchmarks passed · **Cost across the matrix:** $2.98

## Strengths
- There aren't any to report on this rubric. Every benchmark ended with no parseable JSON.

## Watch-outs
- Crunchbase run was $1.91 with 9 tool calls and no output. This is what "burned for nothing" looks like — almost all of the cost was LLM thinking tokens, none of it produced an answer.
- Reddit, CX, Coding, and Web Scraping all ended with empty final messages after between 1 and 22 tool calls. The model engaged with the agent loop, then never wrote the answer.
- This is the tool-loop pathology at its most expensive. Total spend was $2.98 — second-most-expensive model in the matrix — for the worst pass rate.
- We tested the production endpoint as of November 2026. If Qwen updates the weights or the serving stack, this profile may stop being accurate; rerun before you make any decisions on it.