Qwen3.7 Flash ranks #114 of 228 on the public leaderboard with a score of 47.04/100. It does not yet have enough sourced coverage for a verified position.
Qwen3.7 Flash is a proprietary model with a 1M context window. It uses an explicit reasoning mode, which can improve complex problem solving while adding latency and token use.
Available from QwenCloud as qwen3.7-flash and through OpenRouter as qwen/qwen3.7-flash. The APIs expose switchable thinking, function calling, built-in tools, structured output, and prompt caching.
OpenRouter lists qwen/qwen3.7-flash as released on July 27, 2026. QwenCloud's first-party documentation confirms text, image, and video input, text output, a 1,000,000-token context window, up to 65,536 output tokens, a 256,000-token thinking budget, function calling, built-in tools, and structured output. We found no exact public benchmark table for the release, so the model remains outside the rankings.
1 of 766 tracked benchmark slots currently have displayable evidence. Missing categories stay blank.
Model research
What Qwen3.7 Flash is built for
Qwen3.7 Flash is the low-cost model in Alibaba’s Qwen3.7 API family. It accepts text, images, and video, then produces text. QwenCloud documents switchable thinking, function calling, built-in tools, and structured output, which makes the model a candidate for agent steps, visual document work, screen understanding, and general chat applications.
The technical limits are unusually broad for the price tier: a 1,000,000-token context window, up to 65,536 output tokens, and a 256,000-token thinking budget. QwenCloud also allows up to 256 image URLs, 250 Base64 images, or 64 videos in one request, all within the shared context limit.
The price changes with prompt length
QwenCloud’s first-party rate starts at $0.03 per million input tokens and $0.13 per million output tokens for requests up to 32K tokens. The next tier costs $0.10 input and $0.40 output above 32K through 256K. Requests above 256K through 1M cost $0.20 input and $0.80 output.
The pricing card shows the lowest tier so it stays comparable with other headline API rates. Long-context cost should use the tier that matches each request, not the $0.03/$0.13 entry rate. OpenRouter also serves the model, but provider routing, caching, and account terms can change the effective bill.
What the public evidence does not establish
The current sources establish the API shape, modalities, context limits, tools, and price schedule. None of them includes a benchmark table. The one result on this card comes from Epoch AI, which ran Qwen3.7 Flash on the 285 private FrontierMath Tiers 1–3 problems and published 19.3%. One math result can’t rank a model, so we leave the other categories blank instead of borrowing results from Qwen3.7 Plus, Qwen3.7 Max, or an earlier Flash model.
OpenRouter dates its listing to July 27, 2026, while QwenCloud exposes a dated qwen3.7-flash-2026-07-15 snapshot. A snapshot identifier is not necessarily a public launch date, so the profile uses OpenRouter’s availability date and keeps the first-party snapshot only as technical documentation. The next useful update is another sourceable evaluation, not an inferred family score.