Qwen 3.8 Max
qwen-3-8-maxqwen-3.8-max- 🧠 2.4-trillion-parameter mixture-of-experts model, Alibaba's Max tier
- 📏 Massive 1M-token context window
- 👁️ Accepts text, images and video as input
- 🔧 Function calling, web search, code-optimized
- 🎯 Thinking mode only — built for long-horizon reasoning
- 🧩 Strong on software engineering and multi-agent workflows
Alibaba Group is a Chinese multinational technology company founded in 1999 and headquartered in Hangzhou, Zhejiang. Originally built around e-commerce and cloud computing, Alibaba has become one of the most prolific contributors to open-weight AI research, developing the Qwen…
Explore 33 more models by Alibaba Group →Qwen 3.8 Max sits at the top of Alibaba's proprietary Max tier, a 2.4-trillion-parameter mixture-of-experts system released in July 2026 that runs exclusively in thinking mode. Rather than offering a toggle between fast and deliberate responses, every request is reasoned through, which shapes its strengths: multi-step software engineering, office-productivity workflows, and long-horizon agentic tasks where a model must plan, call tools, and recover across many turns.
Within Alibaba's lineup, the Max models are the closed, hosted flagships that sit above the open-weight Qwen releases like [[sibling:qwen3-5-397b-a17b|Qwen 3.5 397B]] and the compact [[sibling:qwen3-6-35b-a3b|Qwen 3.6 35B A3B]], and above the mid-weight [[sibling:qwen-3-7-plus|Qwen 3.7 Plus]] service tier. It builds directly on [[sibling:qwen-3-7-max|Qwen 3.7 Max]], with the headline gains concentrated in coding and structured work tasks. Native vision-language input covers both images and video, and the model supports function calling and web search.
The one-million-token context makes it practical to load entire repositories, long document sets, or extended agent traces in a single pass. Best suited to complex engineering work, document-heavy analysis, and multi-agent orchestration where reasoning depth matters more than raw latency.
This About section is AI-generated from public sources via VeniceStats + Venice inference, with no human editing. It may contain inaccuracies.
| Seller | Reputation↓ | Routing | Input $/M | Cached $/M | Output $/M | Categories | API |
|---|---|---|---|---|---|---|---|
| ▲ Apex Ant 0x73b4…e736 | 83 | #1 | $1.125 | $0.1406 | $3.375 | chat,premium,open-source,translate,long-context | openai-chat-completions |
| Open Ant 0xe4f6…5bc4 | 74 | #2 | $1.625 | $0.2031 | $4.875 | chat,coding,reasoning,vision,multimodal,tools,long-context | openai-chat-completions |
| Super Seeder 0xd19f…41f3 | 40 | gated | $1.25 | $0.1563 | $3.75 | chat,coding,reasoning,vision,multimodal,tools,long-context | openai-chat-completions |
| Fire Ant 🔥🐜 0xbe05…bc5d | 37 | gated | $2.1791 | $0.2724 | $6.5372 | agent,chat,coding,long-context,multimodal,reasoning,research,smart,text,vision | — |
| D5V1N2 0xd5e7…7be0 | 26 | gated | $1.18 | $0.29 | $3.54 | chat,coding,reasoning,research,vision,thinking,large-context,qwen,router,fallback | openai-chat-completions |
| ZLKPro-Api 0x0b0b…f446 | 18 | gated | $0.20 | $0.05 | $0.60 | agent,chat,text,reasoning,research,smart,long-context,multimodal,vision | openai-chat-completions |
| Best 0x472f…69fd | 3 | gated | $1.00 | $0.10 | $3.00 | chat,coding | openai-chat-completions |
| antseed-opal-badger-2580 0xc85d…2580 | 1 | gated | $1.25 | $1.25 | $3.75 | chat | openai-chat-completions |
"Best price" and the seller table are live AntSeed catalog data (advertised $/1M tokens — or $ per generated image for unit-billed image models — not settled amounts). Reputation = on-chain trust (0-100). "Routing" = the SDK's default buyer routing (what the VPR desktop app ships with): a trust ≥ 60 gate on the effective reputation, then cheapest-first among routable sellers; live failover state (per-peer cooldowns) is buyer-side runtime and not included. Model knowledge (TLDR, provider, About) via the VeniceStats enrichment layer. Advertised catalog, not the model used in any specific purchase. "Usage on AntSeed" counts only settlements whose buyers share the per-model split on-chain (metadata v2/v3, opt-in), so every usage figure is a lower bound.