Qwen3.8-Omni-Flash
Released
2026-09-18
Last refreshed
2026-10-09
Status
Researched 45d ago
ProprietaryCommercial use: restrictedMultimodalCodingRAGAgentsLong contextVisionJSON / Tool useMultimodalAgents
Alibaba releases · 53 in the last 12 months · this family litChangelog →
Specifications
- Family
- Qwen3.8
- Released
- 2026-09-18
- Context
- 1M
- Max output
- 128K
- Specialization
- general
- Openness
- Proprietary
- License
- ProprietaryRestricted
- Weights
- Not released
- Code
- Unknown
- Training
- Pretrained
Created by
Pricing
Output / 1M
$0.470
Input / 1M
$0.150
Cheapest of 2 routes · Alibaba Cloud PAI-EAS · cache read $0.016
Links
About
Qwen3.8-Omni-Flash is Qwen's hosted omni-modal model for agentic work, accepting text, image, audio, and video input and returning text only, with a context of up to 1M tokens (991K max input, 131K max output). It handles mixed audio-video workflows such as video editing, narration, and multimedia summarization, understands two- and four-channel spatial audio, and is served through QwenCloud with DashScope- and OpenAI-compatible APIs under the model id `qwen3.8-omni-flash`.
Top use-case fit: coding, agents, and build tasks
Coding
Q/$ BProviders(2)
Compare all 2Compare API pricing across 2 tracked provider routes for input and output tokens, batch, and cached reads when available.
| Provider | Input / 1M | Output / 1M | Cache | Route |
|---|---|---|---|---|
| Alibaba Cloud PAI-EAS | $0.150 | $0.470 | read $0.016 | Serverless |
| OpenRouter | $0.150 | $0.470 | read $0.016 | Serverless |
Capabilities
VisionMultimodalReasoningJSON / Tool useStructured OutputsPrompt CachingAudio
Benchmark peer barsfor Coding
SWE-bench ProRank 13 of 60
LiveCodeBenchRank 2 of 78
Benchmark scores(7)
Scores are benchmark-specific and are direction-aware: the same numeric gap can mean very different outcomes across suites. Use the leaderboard context and this model's provider route to decide whether the winning margin is meaningful for your workload.
| Benchmark | Score | Version | Evaluation | Source |
|---|---|---|---|---|
| 63.3 | SWE-bench Pro (Qwen-refined task set)Observed 2026-10-26 | Configuration: Qwen3.8-Omni-Flash Harness: Claude Code (temperature 1.0, top_p 0.95, 256K context) Evaluator: Qwen (Alibaba) Confidence: self-reported Notes: Qwen Team tech report 'Qwen3.8-Omni: Towards Native Omni-Modal Agents' (arXiv 2609.25611v1, 2026-09-22), Table 2 column 'Qwen3.8-Omni-Flash', row 'SWE-bench Pro': 63.3. Footnote 2: problematic tasks corrected and models re-evaluated on the refined benchmark (comparability caveat). Retrieved 2026-10-26. | ||
| 80.5 | SWE-bench MultilingualObserved 2026-10-26 | Configuration: Qwen3.8-Omni-Flash Harness: mini-SWE-agent (temperature 1.0, top_p 0.95, 256K context) Evaluator: Qwen (Alibaba) Confidence: self-reported Notes: Qwen Team tech report 'Qwen3.8-Omni: Towards Native Omni-Modal Agents' (arXiv 2609.25611v1, 2026-09-22), Table 2 column 'Qwen3.8-Omni-Flash', row 'SWE-bench Multilingual': 80.5. Retrieved 2026-10-26. | ||
| 48.9 | NL2Repo-BenchObserved 2026-10-26 | Configuration: Qwen3.8-Omni-Flash Harness: Claude Code (repo-fetching Bash commands disabled) Evaluator: Qwen (Alibaba) Confidence: self-reported Notes: Qwen Team tech report 'Qwen3.8-Omni: Towards Native Omni-Modal Agents' (arXiv 2609.25611v1, 2026-09-22), Table 2 column 'Qwen3.8-Omni-Flash', row 'NL2Repo-Bench': 48.9. Footnote 4: Bash commands that fetch the target repository disabled to prevent reward hacking. Retrieved 2026-10-26. | ||
| 91.0 | GPQA DiamondObserved 2026-10-26 | Configuration: Qwen3.8-Omni-Flash Evaluator: Qwen (Alibaba) Confidence: self-reported Notes: Qwen Team tech report 'Qwen3.8-Omni: Towards Native Omni-Modal Agents' (arXiv 2609.25611v1, 2026-09-22), Table 2 column 'Qwen3.8-Omni-Flash', row 'GPQA Diamond': 91.0. Retrieved 2026-10-26. | ||
| 36.5 | HLE (tool setting not stated; GPT-4o judge)Observed 2026-10-26 | Configuration: tool setting not stated; GPT-4o judge Evaluator: Qwen (Alibaba) Confidence: self-reported Notes: Qwen Team tech report 'Qwen3.8-Omni: Towards Native Omni-Modal Agents' (arXiv 2609.25611v1, 2026-09-22), Table 2 column 'Qwen3.8-Omni-Flash', row 'HLE': 36.5. Tool setting not stated, so filed under generic hle. Retrieved 2026-10-26. | ||
| 92.6 | LiveCodeBench v6Observed 2026-10-26 | Configuration: Qwen3.8-Omni-Flash Evaluator: Qwen (Alibaba) Confidence: self-reported Notes: Qwen Team tech report 'Qwen3.8-Omni: Towards Native Omni-Modal Agents' (arXiv 2609.25611v1, 2026-09-22), Table 2 column 'Qwen3.8-Omni-Flash', row 'LiveCodeBench v6': 92.6. Retrieved 2026-10-26. | ||
| 83.5 | CharXiv RQ (reasoning), without code interpreterObserved 2026-10-26 | Configuration: Qwen3.8-Omni-Flash Harness: Static eval, no code interpreter Evaluator: Qwen (Alibaba) Confidence: self-reported Notes: Qwen Team tech report 'Qwen3.8-Omni: Towards Native Omni-Modal Agents' (arXiv 2609.25611v1, 2026-09-22), Table 3 column 'Qwen3.8-Omni-Flash', row 'CharXiv (RQ)' without CI: 83.5 (w/ CI row not packed). Fixed step-by-step \boxed{} prompt for Qwen models (footnote 3). Retrieved 2026-10-26. |
Migration checks
No linked migration route is available for this model yet.
API versions
qwen3.8-omni-flashQwen3.8-Omni-FlashRankings & picks(1)
Created by
Pricing
Output / 1M
$0.470
Input / 1M
$0.150
Cheapest of 2 routes · Alibaba Cloud PAI-EAS · cache read $0.016
Links