Alibaba Qwen 3.6 Flash — fast tier. 1M context.
Qwen3.6-Flash is the lightweight Flash tier of Alibaba's Qwen3.6 generation, with 1M context, thinking mode, function calling, built-in tools, and structured output documented upstream. Since the Oct 15, 2026 update of Alibaba Cloud Model Studio's text-generation guide, Alibaba lists qwen3.6-flash (snapshot qwen3.6-flash-2026-04-16) under legacy models that are no longer recommended. Legacy is not retirement: the model is still served, and it remains available on TheRouter as qwen/qwen3.6-flash. For new text workloads, Alibaba's recommended low-cost tier is qwen3.8-flash (qwen/qwen3.8-flash); for image or video input, use qwen/qwen3.7-flash.
TheRouter exposes the model as qwen/qwen3.6-flash through the standard OpenAI-compatible Chat Completions path. Runtime configuration comes from standard-models.yaml: 1,000,000-token context, 32,768 max completion tokens, text input and output, and support for temperature, max_tokens, top_p, tools, tool_choice, response_format, and stop. This curated page does not override live routing fields; it documents the source trail and production placement around the existing route.
Existing deployments can keep Qwen3.6-Flash as a cost-control lane for high-volume chat, extraction, classification, summarization, and office automation where tool calling or JSON output matters. New deployments should start on qwen/qwen3.8-flash, and existing traffic should be regression-tested against it before switching. For harder reasoning, code-agent, or quality-sensitive requests, Alibaba's current recommendations are qwen3.7-plus for balanced cost and qwen3.8-max for the strongest reasoning.
Alibaba Cloud Model Studio lists qwen3.6-flash as a Qwen3.6 model with 1M context, thinking mode, function calling, built-in tools, and structured output. The upstream pricing card is tiered at 256K input tokens for international usage.
TheRouter serves qwen/qwen3.6-flash through the OpenAI-compatible chat route as text-in/text-out with 1,000,000 catalog context, 32,768 max output, and supported parameters temperature, max_tokens, top_p, tools, tool_choice, response_format, and stop. The current TheRouter catalog does not expose Alibaba's enable_thinking control as a public reasoning parameter for this route.
| Type | Rate |
|---|---|
| Prompts up to 256K tokens | |
| Input | $0.270 / 1M tokens |
| Output | $1.62 / 1M tokens |
| Cached input | $0.270 / 1M tokens |
| Prompts over 256K tokens | |
| Input | $1.08 / 1M tokens |
| Output | $4.32 / 1M tokens |
| Cached input | $1.08 / 1M tokens |
Tiered pricing: when a request's prompt (input, cached tokens included) is over 256K tokens, every token in that request is billed at the long-context rates; exactly 256K is still the base rate. This model publishes no separate cached-input price: cached tokens are billed at the input rate of their group (the "Cached input" row shows that rate).
| Context window | 1,000,000 tokensalibabacloud.com ↗ | verified |
| Thinking mode | Alibaba lists qwen3.6-flash as supporting thinking mode; TheRouter's current public parameter list for this route does not expose a reasoning parameteralibabacloud.com ↗ | verified |
| Tooling | Function calling, built-in tools, and structured output are listed as supported upstream; TheRouter exposes tools, tool_choice, and response_formatalibabacloud.com ↗ | verified |
| Modalities | Text input; text output on TheRouter | verified |
| Maximum completion | 32,768 tokens in standard-models.yaml | verified |
| Upstream list pricing — International ≤256K | $0.25 input / $1.50 output per 1M tokensalibabacloud.com ↗ | verified |
| Upstream list pricing — International 256K–1M | $1.00 input / $4.00 output per 1M tokensalibabacloud.com ↗ | verified |
| Training cutoff | Not publicly disclosed | unknown |
| License / weights | No public open-weight release found in reviewed Alibaba Cloud sources | unknown |
| Benchmark | Distribution | Score | Source |
|---|---|---|---|
Official benchmark table The reviewed Alibaba Cloud Model Studio pages position qwen3.6-flash by capabilities, context, and pricing, but do not publish a general benchmark table for this endpoint. Treat production evals as required before replacing Plus or Max traffic. | — | Not publicly disclosed | — |
Coding benchmark table For coding tools, Alibaba recommends qwen3.7-plus, with qwen3.8-max for the strongest reasoning and qwen3.8-flash as the lower-cost switch; qwen3.6-flash is listed under legacy models. No SWE-bench, LiveCodeBench, or agentic-coding score for qwen3.6-flash was found in the free public sources used here. | — | Not publicly disclosed | — |
Latency benchmark table Flash positioning implies a faster, cheaper serving lane, but this pass found no official TTFT, tokens/sec, or latency distribution for qwen3.6-flash. Measure your own request mix before publishing latency promises. | — | Not publicly disclosed | — |
Use the global api.therouter.ai endpoint shown below for new integrations; the legacy China accelerated endpoint is retired.
curl https://api.therouter.ai/v1/chat/completions -H "Content-Type: application/json" -H "Authorization: Bearer $THE_ROUTER_API_KEY" -d '{
"model": "qwen/qwen3.6-flash",
"messages": [
{"role": "user", "content": "Summarize the key points from this input."}
]
}'Use Qwen3.6-Flash through TheRouter's OpenAI-compatible Chat Completions endpoint for low-cost long-context text workloads.
curl https://api.therouter.ai/v1/chat/completions \
-H "Authorization: Bearer $THEROUTER_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "qwen/qwen3.6-flash",
"messages": [
{"role": "system", "content": "Return concise JSON."},
{"role": "user", "content": "Classify these support tickets by urgency and product area."}
],
"temperature": 0.1,
"top_p": 0.8,
"max_tokens": 1200,
"response_format": {"type": "json_object"}
}'The Oct 15, 2026 update of the Model Studio text-generation guide lists qwen3.6-flash, together with qwen3.6-plus, qwen3.7-max, and the Qwen3.5 models, as legacy and no longer recommended. The Qwen entries in the recommended table are now qwen3.8-max (plus its qwen3.8-max-0902 snapshot), qwen3.8-flash, qwen3.7-plus, and qwen3.7-flash. qwen/qwen3.6-flash stays available on TheRouter.
In August 2026 the Model Studio text-generation guide named qwen3.6-flash as the lightweight Qwen option for reducing costs while preserving 1M context, thinking mode, function calling, built-in tools, and structured output. That positioning was superseded by the Oct 15, 2026 update.
No longer recommended upstream. Since the Oct 15, 2026 docs update, Alibaba lists qwen3.6-flash under legacy models; it is still served and stays available on TheRouter, so existing integrations keep working. For new text workloads use qwen/qwen3.8-flash, Alibaba's recommended low-cost tier; for image or video input use qwen/qwen3.7-flash; for balanced quality, including coding tools, use qwen/qwen3.7-plus.
Yes. Alibaba lists function calling, built-in tools, and structured output for qwen3.6-flash, and TheRouter's route exposes tools, tool_choice, and response_format in the supported parameter list. Validate exact schema behavior in your integration before relying on strict downstream parsers.
Alibaba's international pricing card lists qwen3.6-flash at $0.25 input and $1.50 output per 1M tokens up to 256K input tokens, then $1.00 input and $4.00 output from 256K to 1M. TheRouter may apply its own routing price; use the live model page and your invoice as billing source of truth.
No public open-weight checkpoint was found in the Alibaba Cloud sources reviewed for this pass. Treat qwen/qwen3.6-flash as an API route unless Alibaba publishes a separate downloadable model card.
| source | URL | retrieved | |
|---|---|---|---|
| Context window | alibabacloud.com ↗ | 2026-08-10 | verified |
| Thinking mode | alibabacloud.com ↗ | 2026-08-10 | verified |
| Tooling | alibabacloud.com ↗ | 2026-08-10 | verified |
| Modalities | — | — | verified |
| Maximum completion | — | — | verified |
| Upstream list pricing — International ≤256K | alibabacloud.com ↗ | 2026-08-10 | verified |
| Upstream list pricing — International 256K–1M | alibabacloud.com ↗ | 2026-08-10 | verified |
| Training cutoff | — | — | unknown |
| License / weights | — | — | unknown |
| Official benchmark table | alibabacloud.com ↗ | 2026-08-10 | unknown |
| Coding benchmark table | alibabacloud.com ↗ | 2026-10-19 | unknown |
| Latency benchmark table | alibabacloud.com ↗ | 2026-08-10 | unknown |
| Alibaba moves Qwen3.6-Flash to its legacy model list | alibabacloud.com ↗ | 2026-10-19 | verified |
| Alibaba Cloud Model Studio positioned Qwen3.6-Flash as its low-cost lane (August 2026) | alibabacloud.com ↗ | 2026-08-10 | verified |
| Is qwen/qwen3.6-flash still recommended, and what should I use instead? | alibabacloud.com ↗ | 2026-10-19 | to verify |
| Does qwen/qwen3.6-flash support tool calling and JSON output? | alibabacloud.com ↗ | 2026-08-10 | to verify |
| How is Qwen3.6-Flash priced upstream? | alibabacloud.com ↗ | 2026-08-10 | to verify |
| Is Qwen3.6-Flash open source? | alibabacloud.com ↗ | 2026-08-10 | to verify |