Qwen 3.6 Plus Preview claims 158 t/s on BridgeBench and free access on OpenRouter

• Early third-party testing says Qwen 3.6 Plus Preview combines a 1M-token context window with unusually high speed: BridgeMind AI’s review thread says it hit "158 t/s on BridgeBench" and could one-shot tasks that "paid frontier models struggle with."
• The same early review from BridgeMind AI says the model is "Fast" and "capable on hard tasks" but still "weak on UI design," with a separate design-benchmark post describing the output as "clean" and "functional" rather than standout.
• Availability surfaced quickly through third-party tooling: Wes Roth’s OpenRouter post said Alibaba’s Qwen team had pushed the preview live on OpenRouter, and Teknium’s Hermes Agent note said it was free in Hermes Agent via OpenRouter and on Nous Portal.
• The most concrete deployment caveat in the early coverage comes from the review, which warns engineers to "read the fine print" because prompt and completion handling matters more than raw speed or zero-priced tokens.

BridgeMind AI’s early hands-on review thread frames Qwen 3.6 Plus Preview as a strong coding-oriented preview rather than a broadly polished frontier model. The concrete claims are a 1M context window, "158 t/s on BridgeBench," and performance that in the reviewer’s words could "one-shot" tasks some paid models still miss. The attached [vid:0|BridgeBench run] shows the benchmark pass and the speed comparison screen.

The same thread is also clear about limits. BridgeMind calls it "weak on UI design," and the separate design-benchmark post says the design results were "clean" and "functional" but "not standout." That makes the early signal narrower than a general-purpose model win: strong first impressions on hard code tasks, weaker visual or front-end taste.

The most operationally relevant caveat is trust, not throughput. BridgeMind’s review thread says the model is free on input and output but still ends with "I don't recommend it," arguing that "speed and price don't matter" if engineers are unsure where code and completions are going.

The rollout appears to be happening through aggregators and agent tooling before a broader formal launch narrative. Wes Roth’s post says Qwen 3.6 Plus Preview was pushed live on OpenRouter, while Teknium’s availability post says it is "free on Nous Portal and OpenRouter in Hermes Agent right now."

That matters because it turns the preview into something engineers can test inside existing workflows instead of waiting for a separate platform integration. Based on the two posts, the immediate path is not a new standalone SDK but access through OpenRouter-backed tooling, with Hermes Agent and Nous Portal acting as the quickest surfaced endpoints for hands-on evaluation.