Short answer: not yet. On August 3, 2026, Alibaba made its new flagship Qwen3.8-Max generally available (2.4T total / 95B active parameters, 1M token context) and tagged it "Open-Source" on qwen.ai — but as of August 4 there is no Hugging Face or ModelScope repository, no license, and no confirmed ship date. For developers evaluating frontier model selection and Agent workflow costs, this guide covers: the July–August release timeline, official specs and benchmark tables, MoE architecture breakdown, head-to-head vs Kimi K3 / DeepSeek V4 / Claude, the open-source label controversy, industry context, a six-step evaluation checklist, and FAQ. Key signal: Arena Text Arena #5 (1,496 pts, Preliminary) is the only third-party data point so far — every other benchmark is vendor-run, and the promised open weights have not landed.
July–August 2026 is a dense release window for Chinese trillion-parameter models. The timeline below maps confirmed events and official statements so you can benchmark Qwen3.8-Max against Kimi K3, DeepSeek V4-Flash, and the rest.
| Date | Event |
|---|---|
| July 16 | Moonshot AI releases Kimi K3 — 2.8T parameters (16 of 896 experts activated) — positioning around independent benchmarks and a published technical report |
| July 19 | Qwen3.8-Max preview opens via Token Plan / Qoder / QoderWork at 10% of eventual pricing; no active-parameter count, no benchmark table, and terms of service banning automated production use |
| July 27 | Kimi K3 ships open weights on schedule on Hugging Face, plus attention kernels and an MoE communication library |
| July 31 | DeepSeek ships V4-Flash GA — beats its own V4-Pro preview on nine agentic and coding benchmarks without increasing parameter count |
| August 3 | Qwen3.8-Max goes GA with a full benchmark table and companion agent product "Qwen Office"; Alibaba Hong Kong shares rose ~7%, US shares ~4.5% |
| Expected ~August 10 | Open weights for Qwen3.8-Max and smaller Qwen3.8-27B promised on Hugging Face and ModelScope — as of publication, no repository, license, or firm date exists |
Treating the "Open-Source" tag as shipped weights: qwen.ai marked GA day as open-source, but Hugging Face / ModelScope have no repository or license yet
Total params are not activation cost: 2.4T total with only 95B active — inference bills track the active count, closer to a hundred-billion-class model
Benchmark tables are vendor-run: PaperBench, RecreationBench, and others come from Alibaba's own harness; no third party has reproduced GA numbers
Arena ranking is Preliminary: #5 at 1,496 points on Text Arena is tagged preliminary — not a final verdict
Preview transparency was thin: July preview disclosed no active-parameter count, model card, or safety evaluation; several evaluators advised against production migration
Competitor footnotes carry their own bias: a table footnote implies Fable 5 scores may involve fallbacks — while Alibaba's own methodology is equally unpublished
Verification note: Data in this article is current as of August 4, 2026. Open-weight ship dates, Arena final rankings, and third-party reproductions may change — confirm against official Alibaba announcements and independent evaluators before migrating production workloads.
| Spec | Qwen3.8-Max |
|---|---|
| GA date | August 3, 2026 |
| Total / active parameters | 2.4T / 95B |
| Architecture | Sparse MoE + hybrid attention, built on Qwen3.5 base |
| Context window | 1M tokens (~983K with thinking enabled; 131K max output) |
| Input modalities | Text, image, video |
| API pricing | $2 / $6 per million input/output tokens (implicit cache $0.25, explicit cache write $2.50, read $0.17) |
| Arena Text Arena (Aug 1 snapshot) | #5 overall, 1,496 points (Preliminary) — only non-Anthropic model in top 8 |
| Arena Vision Arena | #2, behind Claude Fable 5 |
| PaperBench (Alibaba-run) | 93.0 (+28.2 vs. prior generation) |
| OSWorld-Verified (Alibaba-run) | 86.1 |
| SWE-bench Pro (Alibaba-run) | 67.7 — behind Fable 5's 80.0 |
| HLE (Alibaba-run) | 43.6 — weakest among flagships compared; Fable 5 posts 53.3 |
| Open weights | Promised "next week"; not live as of publication |
Note: Rows marked "Alibaba-run" come from official launch materials. As of publication, Artificial Analysis, Arena.ai, and other neutral platforms have not reproduced GA-stage scores.
Qwen3.8-Max keeps the Qwen3.5 architectural base and pushes total parameters to 2.4 trillion while activating only 95 billion per token. Inference cost tracks the active count, not the total — which is why Alibaba can price the API at $2/$6 per million tokens, well under Claude Opus 5 ($5/$25) and Claude Fable 5 ($10/$50). This is a bet on architectural efficiency as a pricing lever, not raw scale as a capability lever.
Three tiers — low, medium, xhigh (default) — let developers trade latency for depth. Exposed through enable_thinking on the native API and a reasoning.effort field on the Anthropic-compatible interface.
Showcase cases include a 16-day unsupervised coding project, a 500-plus-step chip-design optimization, and RecreationBench — rebuilding a real application from black-box interaction and visual feedback only. A partial trace is public on GitHub (qwen-code-dev-bot/oh-my-cli), but it is not an independently audited, fully reproducible result.
Qwen3.8-Max plugs into Alibaba's "Qwen Office" agent product. The API supports both OpenAI-compatible and Anthropic-compatible protocols — dropping into Claude Code, Codex, Qoder CLI, Qwen Code, and OpenClaw with a base-URL swap.
Separate API from weights: API is live via QwenCloud today; for local deployment, wait for Qwen3.8-27B open weights or confirm Max license terms
Establish a cost baseline: model monthly spend at $2/$6 with $0.25 cache hits against Kimi K3 at $3/$15
Configure reasoning_effort: start with medium/low for latency-sensitive tasks; reserve xhigh for complex Agent workloads
Run real workload A/B tests: do not migrate on vendor benchmarks alone — blind-review against Kimi K3 on your own codebase
Track third-party reproductions: watch Artificial Analysis, Arena final rankings, and Hugging Face weight announcements
Review compliance and data residency: enterprise scenarios need Alibaba Cloud terms, log retention, and output audit policies assessed
| Model | Lab | Total / active params | Context | Price (in/out per 1M tokens) | Open weights? | Independent benchmark |
|---|---|---|---|---|---|---|
| Qwen3.8-Max | Alibaba | 2.4T / 95B | 1M | $2 / $6 | Promised, not shipped | None yet |
| Kimi K3 | Moonshot AI | 2.8T / ~50B (16 of 896 experts) | ~1.05M | $3 / $15 | Shipped July 27 | AA Index ~57.11 |
| DeepSeek V4-Pro | DeepSeek | 1.6T / 49B | 1M | Not fully published | Shipped | 80.6% on SWE-bench Verified |
| DeepSeek V4-Flash | DeepSeek | Same as V4-Pro | 1M | Not fully published | Shipped | Beats V4-Pro on 9 agentic/coding benchmarks |
| Claude Opus 5 | Anthropic | Undisclosed | 1M | $5 / $25 | Closed | Top-tier Arena ranking |
| Claude Fable 5 | Anthropic | Undisclosed | 1M | $10 / $50 | Closed | #1 on Arena Text Arena |
Easy to miss: Kimi K3 disclosed roughly 50 billion active parameters and DeepSeek disclosed 49 billion for V4-Pro — but Alibaba disclosed nothing about active-parameter count during the July preview, only revealing "95B" at GA. In the only apples-to-apples independent test (269-file real software architecture task, blind-reviewed), Kimi K3 scored 83/100 and Qwen3.8-Max preview scored 80/100 — trading blows, not dominating.
Claims like "second only to Fable 5" or "beats GPT-5.6 Sol on most benchmarks" currently rest entirely on Alibaba's own characterization — treat them as vendor claims until weights ship and independent leaderboards catch up.
If you plan to wire Qwen3.8-Max into Claude Code, Qoder, or OpenClaw for long-session Agent workflows, running CLI tools on a laptop or unstable Linux VPS often means memory pressure, dropped sessions, and missing Xcode/Metal toolchains. For production environments that need stable SSH sessions, DerivedData caching, and iOS CI/CD automation, when local Mac capacity falls short, NodeMini Mac Mini cloud rental is usually the better fit — dedicated nodes with second-scale provisioning so Agent and build tasks run on a real Mac continuously. See Mac Mini rental rates.
Sources: Alibaba Cloud official blog and press room, Arena.ai leaderboards (August 1, 2026 snapshot), independent analysis from Apidog, Yotta Labs, TechNode, and Apple Intelligence China coverage. Verify latest official announcements before publishing.
No. The API is live through Alibaba Cloud's Model Studio, but model weights have not been published on Hugging Face or ModelScope. Alibaba's site tags it "Open-Source," but that label currently describes an intent, not a shipped artifact — weights for Qwen3.8-Max and Qwen3.8-27B are promised "next week" with no confirmed date. For deployment environment questions, see Mac Mini rental rates.
There's no single authoritative head-to-head yet. The only independent comparison — a blind-reviewed software architecture task — had Kimi K3 score 83/100 versus Qwen3.8-Max's 80/100, essentially a tie. Kimi K3's advantage is public weights and an Artificial Analysis score; Qwen3.8-Max's advantage is lower API pricing and broader native multimodal support.
For the full model, yes — a 2.4T-parameter checkpoint (even with 95B active) is a multi-node datacenter artifact. The API sidesteps that entirely. For local deployment, the more realistic target is the smaller Qwen3.8-27B, shipping open-weight alongside the flagship and sized for ordinary GPU hardware.
Treat them as a vendor's claim, not a verified result. Every published score comes from Alibaba's own test harness, including several benchmarks Alibaba created (QwenSWEBench, RecreationBench, and others). No independent platform had reproduced the GA-release numbers as of publication, and Alibaba's Arena Text Arena entry is still marked "Preliminary." Wait for third-party reproductions or test against your own workload.
Qwen now powers generative AI features inside Apple Intelligence for users in China, running on-device on recent iPhones. That's a live example of a Chinese open-weight model family becoming embedded infrastructure inside a major global consumer platform — independent of whether you ever call the Qwen API directly. More questions? See the help center.