Qwen3 Models by Alibaba
Last refreshed 2026-10-02. Next refresh: weekly.
Details
Capabilities
About
Qwen3 is Alibaba’s Qwen Team open-weight language-model generation, released as dense sizes from 0.6B to 32B plus MoE models including Qwen3-30B-A3B and flagship Qwen3-235B-A22B. Models support hybrid thinking and non-thinking modes so users can trade depth of reasoning for speed, plus broad multilingual coverage and stronger coding and agent-style tool use than prior Qwen generations. Post-trained and base checkpoints are published for local and server deployment.
Decision facts
- Best fit
- vision and multimodal workreasoningJSON / Tool use
- Capability starting point
- Qwen3-235B-A22B with 128K context and structured outputs
- Lowest tracked input
- Qwen3-8B · $0.035/1M · Novita AI
- Closest related family
- Ovis Image
Current Variants
Use-when guidance is based on each model's tracked capabilities, context window, release date, and replacement status.
Use when the workload needs 128K context, 8B parameters, and structured outputs.
Use when the workload needs 128K context and 8B parameters.
Use when the workload needs 128K context, 235B parameters, and structured outputs.
Use when the workload needs 40K context, 32B parameters, and structured outputs.
Use when the workload needs 128K context, 30B parameters, and structured outputs.
Use when the workload needs 256K context, JSON / Tool use, and structured outputs.
Use when the workload needs 160K context, 8B parameters, and reasoning.
Use when the workload needs 160K context, 671B parameters, and reasoning.
Use when the workload needs 40K context and 600M parameters.
Use when the workload needs 40K context and 1.7B parameters.
Use when the workload needs 40K context, 14B parameters, and structured outputs.
Use when the workload needs 40K context and 4B parameters.
| Model | Use when | Released | Signals | Status |
|---|---|---|---|---|
| Qwen3-Next-80B-A3B | Use when the workload needs structured outputs. | 2025-12 | structured outputs | Current |
| Qwen-Plus | Use when the workload needs 1M context. | 2025-11 | 1M context | Current |
| Qwen3-8B | Use when the workload needs 128K context, 8B parameters, and structured outputs. | 2025-08 | 128K context8B parametersstructured outputs | Current |
| Qwen3-8B-Instruct | Use when the workload needs 128K context and 8B parameters. | 2025-08 | 128K context8B parameters | Current |
| Qwen-Flash | Use when the workload needs 1M context. | 2025-08 | 1M context | Current |
| Qwen3-235B-A22B | Use when the workload needs 128K context, 235B parameters, and structured outputs. | 2025-04 | 128K context235B parametersstructured outputs | Current |
| Qwen3-32B | Use when the workload needs 40K context, 32B parameters, and structured outputs. | 2025-04 | 40K context32B parametersstructured outputs | Current |
| Qwen3-30B-A3B | Use when the workload needs 128K context, 30B parameters, and structured outputs. | 2025-04 | 128K context30B parametersstructured outputs | Current |
| Qwen3-Max | Use when the workload needs 256K context, JSON / Tool use, and structured outputs. | 2025-04 | 256K contextJSON / Tool usestructured outputs | Current |
| DeepSeek R1 0528 Distill Qwen3-8B | Use when the workload needs 160K context, 8B parameters, and reasoning. | 2025-01 | 160K context8B parametersreasoning | Current |
| DeepSeek R1 0528 Qwen3-8B | Use when the workload needs 160K context, 671B parameters, and reasoning. | 2025-01 | 160K context671B parametersreasoning | Current |
| Qwen3-0.6B | Use when the workload needs 40K context and 600M parameters. | 2025-01 | 40K context600M parameters | Current |
| Qwen3-1.7B | Use when the workload needs 40K context and 1.7B parameters. | 2025-01 | 40K context1.7B parameters | Current |
| Qwen3-14B | Use when the workload needs 40K context, 14B parameters, and structured outputs. | 2025-01 | 40K context14B parametersstructured outputs | Current |
| Qwen3-4B | Use when the workload needs 40K context and 4B parameters. | 2025-01 | 40K context4B parameters | Current |
Release Timeline
5 release groupsSpecifications(15 models)
| Model | Released | Context | Parameters | Vision | Multimodal | Reasoning | JSON / Tool use | Structured Outputs |
|---|---|---|---|---|---|---|---|---|
| Qwen3-Next-80B-A3B | 2025-12 | — | 80B (3B active) | No | No | No | No | Yes |
| Qwen-Plus | 2025-11 | 1M | — | No | No | No | No | No |
| Qwen3-8B | 2025-08 | 128K | 8B | No | No | No | No | Yes |
| Qwen3-8B-Instruct | 2025-08 | 128K | 8B | No | No | No | No | No |
| Qwen-Flash | 2025-08 | 1M | — | No | No | No | No | No |
| Qwen3-235B-A22B | 2025-04 | 128K | 235B | No | No | No | No | Yes |
| Qwen3-32B | 2025-04 | 40K | 32B | No | No | No | No | Yes |
| Qwen3-30B-A3B | 2025-04 | 128K | 30B | No | No | No | No | Yes |
| Qwen3-Max | 2025-04 | 256K | — | Yes | Yes | No | Yes | Yes |
| DeepSeek R1 0528 Distill Qwen3-8B | 2025-01 | 160K | 8B | No | No | Yes | No | No |
| DeepSeek R1 0528 Qwen3-8B | 2025-01 | 160K | 671B | No | No | Yes | No | No |
| Qwen3-0.6B | 2025-01 | 40K | 0.6B | No | No | No | No | No |
| Qwen3-1.7B | 2025-01 | 40K | 1.7B | No | No | No | No | No |
| Qwen3-14B | 2025-01 | 40K | 14B | No | No | No | No | Yes |
| Qwen3-4B | 2025-01 | 40K | 4B | No | No | No | No | No |
Available From(11 providers)
Pricing
Popular comparisons in this family
Comparisons
- Qwen3-Max vs GPT-4o (08-06)
- Qwen3-Max vs Claude Sonnet 4.6
- Qwen3-Max vs DeepSeek V4 Pro
- Qwen3-30B-A3B vs Mistral Large 2.1 (2411)
- Qwen3-30B-A3B vs Llama 3.3 70B
- Qwen3.6-Max vs Qwen3-Max
- GPT-5 vs Qwen3-Max
- Claude Opus 4.7 vs Qwen3-Max
Related Model Families
Models(15)
Qwen3-Next-80B-A3B
Qwen-Plus
Qwen3-8B
Qwen3-8B-Instruct
Qwen-Flash
Qwen3-235B-A22B
Qwen3-32B
Qwen3-30B-A3B
Qwen3-Max
DeepSeek R1 0528 Distill Qwen3-8B
DeepSeek R1 0528 Qwen3-8B
Qwen3-0.6B
Qwen3-1.7B
Qwen3-14B
Qwen3-4B