Qwen: Qwen3.5-9B (batch)

The qwen/qwen3.5-9b:batch entry was last present in the OpenRouter snapshot of 2026-09-21 and absent from the snapshots of 2026-09-23 and 2026-09-24. A live fetch on 2026-09-24 finds it still gone. The model catalog lists the standard qwen/qwen3.5-9b entry and no batch variant of it; …/models/qwen/qwen3.5-9b:batch/endpoints answers 200 with "endpoints":[]; and …/model/qwen/qwen3.5-9b:batch answers 404, Model not found: qwen/qwen3.5-9b:batch. OpenRouter's variant guide reads that pair as an unavailable model: "When the resolved catalog entry does not exist … the single-model route returns 404 and the endpoints route returns 200 with "endpoints": []. Treat either as an unavailable model, not as a signal to retry with the base slug." The batch entry is gone from the router.

Qwen3.5-9B itself is still available. Fetched on 2026-09-24, the standard entry lists six endpoints (Darkbloom, SiliconFlow, DeepInfra, Venice, Parasail and Together), and the weights at Qwen/Qwen3.5-9B are public and not gated. Nothing was renamed: the standard entry keeps the qwen/qwen3.5-9b-20260310 canonical slug that the batch row carried, and no batch entry under another Qwen slug took its place. It was not folded into another tier either. The catalog simply has no batch tier for this model any more.

The provider behind the batch entry can be identified. OpenRouter's batch quickstart says "A batch runs on one provider. By default OpenRouter picks the cheapest eligible batch endpoint for the model". Of the six endpoints on the standard entry, only Together's matches every value pinned for the batch row when it went missing: the context length, the completion cap, the same sixteen supported parameters, and both the input and the output rate. No other endpoint charges the batch row's input rate. That rate pair is the one Together's changelog set for Qwen3.5-9B in a pricing update effective May 29, 2026. The pinned values are kept in data/vanished/answered/openrouter-models--qwen-qwen3-5-9b-batch.md.

Together has not withdrawn the model. Its serverless model list, fetched 2026-09-24, still carries Qwen/Qwen3.5-9B with the same context length and the same two rates, and names it the model to "start with" for vision. Its deprecations page mentions Qwen3.5-9B only as the recommended replacement for retired models, among them openai/gpt-oss-20b, removed from serverless on 2026-09-14. Its batch overview still says "Most serverless models support batch processing", and its batch inference page offers batch jobs on "Any serverless model". Together still lists this model for batch inference: the "Batch API price" table on its pricing page, fetched 2026-09-24, has a "Qwen3.5 9B" row at the same input and output rates that were pinned for the vanished OpenRouter row. So the provider that priced this entry still serves the model and still sells batch inference on it.

The reason is therefore unknown. OpenRouter does not publish per-entry reasons for withdrawing a variant, and Together's current batch listing does not explain why OpenRouter dropped its own. It was part of a wave: between the 2026-09-21 and 2026-09-23 snapshots, eight other :batch entries left the catalog with this one, among them minimax/minimax-m3:batch and thinkingmachines/inkling:batch. And batch entries come back: openai/gpt-oss-20b:batch, absent from the 2026-09-21 snapshot, is listed in the 2026-09-23 snapshot and in the live catalog on 2026-09-24.

That is why the feeds: binding stays. If qwen/qwen3.5-9b:batch is ever a catalog entry again, it binds straight back to this page. The timeline dates the first snapshot without the batch entry. It does not record a retirement of Qwen3.5-9B, which is live, open-weight and listed on OpenRouter under six providers.