Qwen3.8 2.4T A95B is an open-weight, text-only Qwen3.8 model available through Novita AI’s serverless API. It is aimed at developers who need long-context coding, research, complex reasoning, or agentic workflows while keeping an open-weight serving option. Novita lists the model ID as qwen/qwen3.8-2.4t-a95b, a 1,000,000-token context window, 131,072 maximum output tokens, and pricing of $2 per 1M input tokens, $0.25 per 1M cache-read tokens, and $6 per 1M output tokens.
Key Takeaways
- Qwen3.8 2.4T A95B is an open-weight sparse MoE model with 2.4T total parameters and 95B active parameters.
- Novita exposes it as a text-only serverless model with function calling, structured outputs, and reasoning support.
- Novita lists a 1,000,000-token context window and a 131,072-token maximum output.
- The model is a better fit for teams that value open-weight portability and text-first workflows than for workloads requiring vision input or a turnkey managed product layer.
Qwen3.8 2.4T A95B Specs and Pricing
| Field | Novita value |
|---|---|
| Display name | Qwen3.8 2.4T A95B |
| Model ID | qwen/qwen3.8-2.4t-a95b |
| Model type | Open-weight sparse MoE |
| Total parameters | 2.4T |
| Active parameters | 95B |
| Context window | 1,000,000 tokens |
| Max output tokens | 131,072 |
| Input modality | Text |
| Output modality | Text |
| Endpoints | chat/completions, responses |
| Features | Serverless, function calling, structured outputs, reasoning |
| Input price | $2 / 1M tokens |
| Cache-read price | $0.25 / 1M tokens |
| Output price | $6 / 1M tokens |
| Rate tier shown | T1: 30 RPM / 50,000,000 TPM |
| Date checked | September 2, 2026 |
Pricing and limits can change. Confirm the Novita AI model page and Novita AI pricing page before making a production cost commitment.
What Qwen3.8 2.4T A95B Can Do
Qwen3.8 2.4T A95B is designed for text-first workloads that benefit from a large context window and strong reasoning capacity. Its open-weight positioning also gives teams a path to evaluate the model against their own serving and deployment requirements rather than treating a managed endpoint as the only option.
The model is suited to repository-scale code review, multi-document research, complex text reasoning, structured generation, and agent loops that retain tool traces or intermediate state. The Qwen model card documents the model weights and configuration files and lists compatibility with serving stacks such as vLLM and SGLang.
When Qwen3.8 2.4T A95B Is a Good Fit
Choose this model when you need:
- open-weight portability and control over the serving layer
- long-context coding, code review, or repository analysis
- text-only research and multi-document reasoning
- agentic workflows using function calling, structured outputs, or reasoning
- a hosted Novita endpoint for evaluation before deeper infrastructure work
The open-weight distinction matters when your team may later self-host, change serving infrastructure, or keep a consistent model artifact across environments. For the managed Qwen3.8-Max model on Novita AI and its product-level access path, see Qwen3.8-Max on Novita AI.
When Qwen3.8 2.4T A95B Is Not the Best Default
This model is not the best default for image or vision input because Novita lists its input modality as text. It may also be more than you need for short classification, lightweight extraction, or high-volume prompts where a smaller model would meet the quality requirement at lower cost.
Choose the managed Qwen3.8-Max path instead when you need its additional product layer, including vision input or a more turnkey managed workflow. If your main question is how to send the first request, use the Qwen3.8-Max API Quick Start rather than this launch and decision page.
Qwen3.8 2.4T A95B API Access on Novita AI
Novita’s model page lists the API model ID as qwen/qwen3.8-2.4t-a95b and supports the chat/completions and responses endpoint families. The Novita OpenAI-compatible API entry point is https://api.novita.ai/openai.
This section intentionally provides access information only. It does not include a first-request walkthrough, complete request parameters, or troubleshooting steps; those belong in a 4.4 quick-start or feature guide.
FAQ
What is the Novita model ID?
Use qwen/qwen3.8-2.4t-a95b.
Is Qwen3.8 2.4T A95B open weight?
Yes. The Qwen model card provides the model weights and configuration files for the post-trained model. Open-weight availability does not mean every deployment has the same hardware, serving, or performance profile.
What context and output limits does Novita list?
Novita lists a 1,000,000-token context window and a 131,072-token maximum output.
What does it cost on Novita AI?
Novita lists $2 per 1M input tokens, $0.25 per 1M cache-read tokens, and $6 per 1M output tokens as checked on September 2, 2026.
Is this the same as Qwen3.8-Max?
It is the open-weight Qwen3.8 2.4T A95B model, while Qwen3.8-Max is the separate managed service path. Compare the verified capabilities and access requirements for your workload before choosing between them.
Sources
- Novita model page: https://novita.ai/models/model-detail/qwen-qwen3.8-2.4t-a95b (checked 2026-09-02)
- Novita pricing page: https://novita.ai/pricing (checked 2026-09-02)
- Qwen model card: https://huggingface.co/Qwen/Qwen3.8-2.4T-A95B (checked 2026-09-02)
- Qwen3.8-Max launch guide: /qwen3-8-max-on-novita-ai-launch-pricing/
- Qwen3.8-Max quick start: /qwen3-8-max-api-quick-start/
