Qwen3.8 2.4T A95BOpen weights
Alibaba · released 2026-08-12
No price from the maker itself is on record for Qwen3.8 2.4T A95B. The lowest of 23 providers is IteraCompute: $1.95 per 1M input tokens and $5.7 per 1M output tokens (as of 2026-11-12). Context window: 262K tokens, up to 131K output tokens per request.
Price by providerUSD per 1M tokens
| Provider | Input | Output | Cached | Notes | |
|---|---|---|---|---|---|
IIteraComputeLowestqwen/ | $1.95 | $5.7 | $0.25 | Lowest | Official docs |
 AIHubMixqwen3.8-2.4t-a95b | $2 | $6 | $0.5 | Official docs | |
EEden AIqwen/ | $2 | $6 | $0.25 | Official docs | |
 Fireworks AIaccounts/ | $2 | $6 | $0.25 | Official docs | |
CCharm Hyperqwen3.8-2.4t-a95b | $2 | $6 | $0.25 | Official docs | |
KKilo Gatewayqwen/ | $2 | $6 | $0.25 | Official docs | |
 DevPass (LLM Gateway)qwen3.8-2.4t-a95b | $2 | $6 | $0.25 | Official docs | |
LLLM Gatewaynovita/ | $2 | $6 | $0.25 | Official docs | |
MModalQwen/ | $2 | $6 | $0.25 | Official docs | |
 Novita AIqwen/ | $2 | $6 | $0.25 | Official docs | |
 OpenRouterqwen/ | $2 | $6 | $0.25 | Official docs | |
RRequestyqwen3.8-2.4T-A95B | $2 | $6 | $0.2 | Official docs |
12 more listings
| Provider | Input | Output | Cached | Notes | |
|---|---|---|---|---|---|
 SiliconFlowQwen/ | $2 | $6 | $0.25 | Official docs | |
 Vercel AI Gatewayalibaba/ | $2 | $6 | $0.25 | Official docs | |
CCortecsqwen3.8-2.4t-a95b | $2.5 | $6 | $0.625 | Official docs | |
OOpperqwen3.8-2.4t-a95b | $2.5 | $6 | $0.63 | Official docs | |
RRequestyqwen3.8-2.4T-A95B@ | $2.5 | $6 | $0.63 | Official docs | |
TTensorXqwen/ | $2.5 | $6 | $0.63 | Official docs | |
 Hugging FaceQwen/ | $2.5 | $6.25 | — | Official docs | |
MMerge Gatewayqwen/ | $2.5 | $6.25 | $0.5 | Official docs | |
 AI/ML APIalibaba/ | $2.645 | $6.6125 | $0.529 | Official docs | |
 Venice AIqwen-3-8-2-4t-a95b | $2.5 | $7.5 | $0.3125 | Official docs | |
 Deep InfraDeprecatedQwen/ | $2 | $6 | $0.2 | Deprecated | Official docs |
RRunInfraDeprecatedInferact/ | $2 | $6 | $0.2 | Deprecated | Official docs |
Providers are those listed by models.dev. This site has not vetted them one by one; each provider's own page has the binding price.
About
Open-weight sparse MoE (2.4T total, 95B active), the open-weight twin of Qwen3.8 Max for coding, research, complex reasoning, and agentic workflows
Description from models.dev.
Specs
- Type
- Text
- Context
- 262K
- Max output
- 131K
- Input
- Text
- Output
- Text
- Reasoning
- Yes
- Tool calling
- Yes
- Structured output
- Yes
- Knowledge cutoff
- —
- Open weights
- Yes
- Released
- 2026-08-12
Source
Prices and specs come from models.dev (MIT licence), updated 2026-11-12. Each provider's name links to its official docs. A provider priced at 0 usually includes the model in a subscription; it is not necessarily a free API.