Qwen3.8 Flash

Aug 2026
Qwen3.8 Flash
Compare in
on
$0.37per task?
$0.15 / $0.47per 1M in / out
55 tok/sfirst token in 2.1 s?
262Kcontext

Rankingin thinking mode?

Full ranking →
1Gemini 4 Argon77.52MiniMax M3.1 Flash72.53Sonnet 5.571.84Muse Wonder Full71.35Opus 5.569.5
···
29Qwen Max (0902)56.230Qwen Flash55.931Fable 554.1

Strong at

S
Agents & work tasks#15 of 175Science#19 of 312
A
Coding#26 of 132Reasoning#52 of 334Long context#48 of 273Vision#21 of 124
B
Knowledge#91 of 318

Alternatives?

ModelScoreCostPer taskWhat changes
CheaperMiniMax M3.1 Flashthinking+30%-41%$0.22Complete model response must be appendedBetterMiMo v2.6 FlashthinkingOpen+15%-83%$0.062Memory (4-bit): 108 GB → 186 GB+4CheaperDeepSeek V4.1 Flash 1017maxOpen+19%-30%$0.26Memory (4-bit): 108 GB → 458 GB+5CheaperGPT-6.1 Solhigh+15%-14%$0.32Remove unsupported parameters+5CheaperGLM 5.3 FlashmaxOpen+8%-24%$0.28Memory (4-bit): 108 GB → 193 GB+5

Run it yourself

estimate
Fits on✗ 8 GB (phone, laptop)✗ 16 GB✗ 24 GB GPU✗ 48 GB GPU✗ 80 GB GPU
Licenceqwen-community-1.0commercial use with conditions

Benchmarks

in Agents & Tool Use
Terminal-Bench 2.186.1#9
Terminal-Bench 4.025.3#26
GDPval-AA (work tasks) 2.156.4#15
AA-LCR (long context) 1.179.7#48
τ³-Banking · AA45.4#11
AutomationBench-AA55.9#30
AA-Briefcase 1.1 · Elo1,579.9#10
DeepSWE 1.1lab58.7
SWE-bench Pro (edition unstated)lab62.5

Watch out 6

max_tokens range · QwenCloud›

With thinking enabled, max_tokens outside 1–32768 returns HTTP 400; use max_completion_tokens to budget reasoning plus answer without that cap. Source ↗

File is not delivered · QwenCloud / Responses›

PDF input through Responses returns HTTP 200 without delivering the file to the model; use Chat Completions or DashScope. Source ↗

Smaller values automatically adjusted · QwenCloud›

In Flash thinking mode, temperature below 0.6 is automatically raised to 0.6. Source ↗

Static YaRN affects shorter texts · Self-hosted›

Static YaRN can hurt short-text performance; change rope_parameters only when longer contexts are required. Source ↗

tool_choice only supports auto or none · QwenCloud›

With thinking enabled, tool_choice supports only auto or none. Source ↗

Include the word JSON · QwenCloud›

JSON Object mode requires the word JSON in a system or user message; otherwise the API returns an error. Source ↗

FAQ

How much does Qwen3.8 Flash cost?›

$0.15 per 1M input tokens and $0.47 per 1M output tokens; cheapest via DeepInfra at $0.11 / $0.38. A typical benchmark task costs about $0.37 at default effort.

Is Qwen3.8 Flash free to use?›

There is no free API; the cheapest paid price is $0.11 per 1M input tokens. The weights are open, so you can run it yourself.

How good is Qwen3.8 Flash, and what is it best at?›

#52 of 334 in Chat & Knowledge, #26 of 132 in Coding, #30 of 133 in Agents & Tool Use, #21 of 124 in Vision & Documents. Strongest at: Coding, Agents & work tasks, Reasoning, Science, Long context, Vision.

Is Qwen3.8 Flash better than Qwen3.8 Max (0902)?›

On AutomationBench-AA, Qwen3.8 Flash scores 55.9 vs 56.2 — about the same. It is 93% cheaper per task.

How can I use Qwen3.8 Flash?›

Through the API of Alibaba, DeepInfra, Featherless, Novita.

What is the context window of Qwen3.8 Flash?›

262K tokens, with up to 131K output tokens.

When was Qwen3.8 Flash released? Is there a newer version?›

Released 2026-08-26. It is the current version in its line.

Is Qwen3.8 Flash open source?›

Yes, the weights are open (qwen-community-1.0 licence).

Can I run Qwen3.8 Flash locally?›

The weights are open and can be self-hosted; hardware needs are not confirmed yet.

Can Qwen3.8 Flash read images?›

It accepts text and images, video as input.

What is Qwen3.8 Flash?›

Qwen3.8 Flash is an open-weights language model from Alibaba (Qwen), released Aug 2026.

What are the best alternatives to Qwen3.8 Flash?›

MiniMax M3.1 Flash (cheaper), MiMo v2.6 Flash (better).